Lex Fridman Podcast - #279 - Alien Debate: Sara Walker and Lee Cronin

I don’t know what it’s like to be an alien.

I would like to know.

Two alien civilizations coexisting on a planet,

what’s that look like exactly?

When you see them and they see you,

you’re assuming they have vision,

they have the ability to construct in 3D and in time.

That’s a lot of assumptions they’re making.

What human level intelligence has done is quite different.

It’s not just that we remember states

that the universe has existed in before,

it’s that we can imagine ones that have never existed

and we can actually make them come into existence.

So you can travel back in time sometimes.


You travel forward in time to travel back.


The following is a conversation

with Sarah Walker and Lee Cronin.

They have each been on this podcast once before

individually and now for their second time,

they’re here together.

Sarah is an astrobiologist and theoretical physicist.

Lee is a chemist and if I may say so,

the real life manifestation of Rick from Rick and Morty.

They both are interested in how life originates

and develops both life here on earth and alien life,

including intelligent alien civilizations

out there in the cosmos.

They are colleagues and friends

who love to explore, disagree and debate

nuanced points about alien life.

And so we’re calling this an alien debate.

Very few questions to me are as fascinating

as what do aliens look like?

How do we recognize them?

How do we talk to them?

And how do we make sense of life here on earth

in the context of all possible life forms

that are out there?

Treating these questions with the seriousness

and rigor they deserve is what I hope to do

with this conversation and future ones like it.

Our world is shrouded in mystery.

We must first be humble to acknowledge this

and then be bold and diving in

and trying to figure things out anyway.

This is the Lex Friedman podcast.

To support it, please check out our sponsors

in the description.

And now, dear friends, here’s Sarah Walker and Lee Cronin.

First of all, welcome back Sarah, welcome back Lee.

You guys, I’m a huge fan of yours.

You’re incredible people.

I should say thank you to Sarah

for wearing really awesome boots.

We’ll probably overlay a picture later on,

but why the hell didn’t you dress up, Lee?

No, I’m just kidding.

This is me dressed up.

You were saying that you’re pink,

that your thing is pink.

My thing is black and white, the simplicity of it.

Where’s the pink?

When did the pink, when did it hit you

that pink is your color?

I became pink about, I don’t know, actually, maybe 2017.

Did you know me when you first?

I think I met you pre pink.

Yeah, yeah, so about 2017, I think.

I just decided I was boring

and I needed to make a statement

and red was too bright, so I went pink, salmon pink.

Well, I think you were always pink.

You just found yourself in 2017.

There’s an amazing photo of him

where there’s like everybody in their black gown

and he’s just wearing the pink pants.

Oh, that was at the Waggonen University.

It’s totally nuts.

100 year anniversary, they got me to give the plenary

and they didn’t find that outfit for me,

so they were all wearing these silly hats and these gowns

and there was me dressed up in pink

looking like a complete idiot.

We’re definitely gonna have to find that picture

and overlay it, big full screen, slow motion.

All right, let’s talk about aliens.

We’ll find places we disagree and places we agree,

life, intelligence, consciousness, universe, all of that.

Let’s start with a tweet from Neil deGrasse Tyson

stating his skepticism about aliens wanting to visit Earth.

Quote, how egocentric of us to think that space aliens

who have mastered interstellar travel across the galaxy

would give, pardon the French,

would give a shit about humans on Earth.

So let me ask you, would aliens care about visiting Earth,

observing, communicating with humans?

Let’s take a perspective of aliens, maybe Sarah first.

Are we interesting in the whole spectrum

of life in the universe?

I’m completely biased, at least as far as I think right now

we’re the most interesting thing in the universe.

So I would expect based on the intrinsic curiosity

that we have and how much I think that’s deeply related

to the physics of what we are,

that other intelligent aliens would want to seek out

examples of the phenomena they are

to understand themselves better.

And I think that’s kind of a natural thing to want to do.

And I don’t think there’s any kind of judgment

on it being a lesser being or not.

It’s like saying you have nothing to learn

by talking to a baby.

You have lots to learn, probably more than you do

talking to somebody that’s 90.

So yeah, so I think they absolutely would.

So whatever the phenomena is that is human,

there will be an inkling of the same kind of phenomena

within alien species and they will be seeking that same.

I think there’s gotta be some features of us

that are universal.

And I think the ones that are most interesting,

and I hope I live in an interesting universe,

are the ones that are driven by our curiosity

and the fact that our intelligence allows us to do things

that the universe wouldn’t be able to do

without things like us existing.

We’re gonna define a lot of terms.

One of them is interesting.


That’s a very interesting term to try to define.

Ali, what do you think?

Are humans interesting for aliens?

Well, let’s take it from our perspective.

We want to go find aliens as a species quite desperately.

So if we put the shoe on the other foot,

of course we’re interesting.

But I’m wondering and assuming

that we’re at the right technological capabilities

to go searching for aliens, then that’s interesting.

So what I mean is,

if there needs to be a massive leap in technology

that we don’t have,

how will aliens prioritize coming to Earth and other places?

But I do think that they would come and find us

because they’d want to find out about our culture,

what things are universal.

I mean, I’m a chemist.

I would say, well, is the chemistry universal, right?

Are the creatures that we’re going to find

making all this commotion,

are they made of the same stuff?

What does their science look like?

Are they off planet yet?

I guess there’s, so I think that Neil deGrasse Tyson

is being slightly pessimistic

and maybe trying to play the tune

that the universe is vast

and it’s not worth them coming here.

I don’t think that,

but I just worry that maybe we don’t have

the ability to talk to them.

We don’t have the universal translator.

We don’t have the right physics,

but sure, they should come.

We are interesting.

I want to know if they exist.

It would make it easier if they just came.

So again, I’m going to use your tweets

like it’s Shakespeare and analyze it.

So Sarah tweeted,

thinking about aliens, thinking about aliens.

So how much do you think aliens

are thinking about other aliens, including humans?

So you said, we humans want to visit.

Like we’re longing to connect with aliens.

Why is that?

Can you introspect that?

Is that an obvious thing that we should be,

like what are we hoping to understand

by meeting aliens exactly?

Asking as an introvert, it’s like,

I ask myself this all the time.

Why go out on a Friday night to meet people?

What are you hoping to find?

I think the curiosity, so when I saw Sarah put that tweet,

I think I answered it actually as well,

which was we are thinking about trying to make contact.

So they almost certainly are,

but maybe there’s a number of classes.

There are those aliens that have not yet made contact

with other aliens like us.

Those aliens have made contact with just one other alien

and maybe it’s an anticlimax and slime, right?

And aliens have made contact

with not just one set of intelligent species, but several.

That must be amazing actually.

Literally there are some place in the universe,

there must be one alien civilization.

Let’s not make contact with not one,

but two other intelligent civilizations.

So they must be thinking about it.

There must be entire degree courses on aliens,

thinking about aliens and universal cultural norms.

Do you think they will survive the meeting?

And by the way, Lee did respond saying,

that’s all the universe wants.

So Sarah said, thinking about aliens, thinking about aliens.

Lee said, that’s all the universe wants.

And then Sarah responded, cheeky universe we live in.

So cheeky is a cheeky version of the word interesting,

all of which we’ll try to define mathematically.

Cheeky might be harder than interesting.

Because there’s humor in that too.


I think there’s a mathematical definition of humor,

but we’ll talk about that in a bit.

Oh interesting.

Yeah, sure there is, yeah.

So if you’re a graduate student alien

looking at multiple alien civilizations,

do you think they survive the encounters?

I think there’s a tendency to anthropomorphize

a lot of the discussions about alien life,

which is a really big challenge.

So usually when I’m trying to think about these problems,

I don’t try to think about us as humans,

but us as an example of phenomenon

that exists in the universe that we have yet to explain.

And it doesn’t seem to be the case

that if I think about the features,

I would argue are most universal about that phenomenon,

that there’s any reason to think

that a first encounter with another lineage

or example of life would be antagonistic.

I think, yeah.

And I think there’s this kind of assumption,

I mean, going back to Neil deGrasse Tyson’s quote,

I mean, it kind of bothers me because there’s a,

I mean, I’m a physicist,

so I know we have a lot of egos

about how much we can describe the world,

but that there’s this like,

because we understand fundamental physics so well,

we understand alien life and we can kind of extrapolate,

and I just think that we don’t.

And the quest there is really, you know,

really to understand something totally new

about the universe, and that thing just happens to be us.

I agree, I agree.

There’s something else more profound.

I think Neil was just being, again,

he’s just trying to stir the pot.

I would say from a contingency point of view,

I want to know how many ways

does the universe build structures, build memories, right?

And then I want to know if those memories

can interact with each other.

And if you have two different origins of life

and then origins of intelligence,

and then these things become conscious,

surely you want to go and talk to them

and figure out what commonalities you share.

And it might be that we’re just unable to conceive

of what they’re going to look like.

They’re just going to be completely different,

you know, infrastructure,

but surely we’ll want to go and find out a map.

And surely curiosity is a property

that evolution has made on earth.

And I can’t see any reason

that it won’t happen elsewhere

because curiosity probably exists

because we want to find innovations in the environment.

We want to use that information to help our technology.

And also curiosity is like planning for the future.

Are they going to fight us?

Are we going to be able to trade with them?

So I think that Neil’s just, I don’t know,

maybe, you know, I mean, give a shit.

That’s really, I think that’s really down on earth, right?

How would aliens categorize humans, do you think?

How would we?

So let’s put it the other way around.

Slime category.

Maybe, no, no, no.

Maybe we could, the thing is a bit odd, right?

Look at Instagram, Twitter,

all these people taking selfies.

I mean, does the universe

is the ultimate state of consciousness,

thinking beings that take photographs themselves

and upload them to an internet

with other thinking beings looking at each other’s photos.

So I think that they will be.

What’s wrong with that?

I did not say there was anything wrong with it.

It’s consciousness manifested at scale.

Selfies, Instagram.

It’s like the mirror test at scale.

Yeah, I do think that curiosity

is really the driving force

for why we have our technology, right?

If we weren’t curious, we wouldn’t go out, left the cave.

So I think that,

so I think that Neil’s got it completely wrong, in fact.

Actually, of course they’d want to come here.

It doesn’t mean they are coming here.

We’ve seen evidence for that.

I guess we can argue about that, right?

But I think that we want,

I desperately, and I know that Sarah does too,

but I won’t speak for you, you’re here.

I desperately want to have missions

to look for life in the solar system right now.

I want to map life over the solar system.

And then I want to understand how we can go

and find life as quickly as possible at the nearest stars,

and also at the same time do it in the lab,

just to compensate, you know?

So, sure.

Yeah, just one more point on this.

If you think about sort of what’s driven

the most features of our own evolution as a species,

and try to map that to alien species,

I always think like optimism is what’s gonna get us furthest.

And so I think a lot of people always think

that it’s like war and conflict is gonna be the way

that alien species will expand out into the cosmos.

But if you just look at how we’re doing it

and how we talk about it,

so is our future in space is always, you know,

built from narratives of optimism.

And so it seems to me that if intelligence does get out

in the universe, that it’s gonna be more optimism

and curiosity driving it than war and conflict,

because those things end up crushing you.

So there might be some selective filter.

Of course, this is me being an optimist.

I’m a half full kind of person, but.

Is it obvious that curiosity, not obvious,

but what do you think?

Is curiosity a more powerful force in the universe

than violence and the will to power?

So, because you said you framed curiosity as a way

to also plan on how to avoid violence,

which is an interesting framing of curiosity.

But I could also argue that violence

is a pretty productive way to operate in the world,

which is like, that’s one way to protect yourself.

The best defense is offense.

I’m not qualified to answer this, but I’ll have a go.

I think violence, let’s not talk about violence.

That’s the summary of this podcast.

I would, yeah, maybe, I would, let’s not call it violence,

but I call it erasure.

So if you think about the way evolution works,

or the way, obviously talk about assembly theory,

but I won’t.

So if you say you build,

curiosity allows you to open up avenues, new graphs, right?

So new features you can play.

The ability to erase those things allows you to start again

and do some pruning.

So the universe, I think curiosity gets you furthest.

Curiosity gets you rockets that land.

It gets you robots that can make drugs.

It gets you poetry and art and communication.

And then, I often think wouldn’t it be great in bureaucracy

to have another world war, not literally a world war now,

please no world war, but the equivalent

so we get remove all the admin bureaucracy, right?

All the admin violence, get rid of it and start again.

Do you know what I mean?

Because you get layers and you get redundant systems built.

So actually a reset, let’s not call it violence,

a reset in some aspects of our culture

and our technology allows us to then build

more important things about the,

because how many cookies do I have to click on?

How many extra clicks do I have in the future of my life

that I could remove and a bit of a reset

would allow us to start again.

And maybe that’s how I suppose our encounter aliens will be.

Maybe they will fight with us and say,

oh, we’re not as excited by you as we thought.

We’ll just get rid of you.

So they might want to reset Earth.

Yeah, why not?

To be like, let’s see how the evolution runs again.

This seems like they’ve, there’s nothing new happening here.

They’re observing for a while.

This is just not, let’s keep it more fun.

Let’s start with the fish again.

I like how you equated violence to resetting your cookies.

I suppose that’s the kind of violence.

In this modern world where words are violence,

resetting cookies is the kind of violence.

I don’t know where that came from.

I’m completely, yeah.

That’s poetic, really.

Okay, so let’s talk about life.

What is life?

What is non life?

What is the line between life and non life?

And maybe at any point,

we can pull in ideas of assembly theory.

Like how do we start to try to define life?

And for people listening,

so Sarah identifies as a physicist

and Lee identifies as a chemist.

Of course, they are very interdisciplinary in nature

in general, but so what is life?


I love asking that question

because it’s so absurdly big.

I know, I love it.

It’s my absolute favorite question in the whole universe.

So I think I have three ways of describing it right now.

And I like to say all three of them

because people latch onto different facets of them.

And so the whole idea of what Lee and I are trying to work on

is not to try to define life,

but to try to find a more fundamental theory

that explains what the phenomenon we call life.

And then it should explain certain attributes

and you end up having a really different framing

than the way people usually talk.

So the way I talk about it three different ways.

Life is how information structures matter

across space and time.

Life is, I don’t know, this one’s from you actually,

simple machines constructing more complex machines.

And the other one is the physics of existence,

so to speak, which is life is the mechanism

the universe has to explore the space of what’s possible.

That’s my favorite.

So can I, yeah, yeah, can I add on to that?

Okay, can you say the physics one again?

Physics of existence.

Yeah, the physics of existence.

I don’t know what to call it.

If you think of all the things that could exist,

only certain things do exist.

And I think life is basically the universe’s mechanism

of bringing things into physically existing

in the moment now.


Yeah, and what’s another one?

So we were debating this the other day.

So if you think about a universe that has nothing in it,

that’s kind of hard to conceive of, right?

Because, and this is where physicists really go wrong.

They think of a universe with nothing in it, they can’t.

And you think.

But nonexistence is really hard to think.

Nonexistence, yeah.

And then you think of a universe with everything in it,

that’s really hard.

And you just have this white blob, right?

It’s just everything.

But the fact we have discrete stuff in the universe,

Beyonce planets, so you’ve got stars, space, planet stuff,

right, the boring stuff, but I would define life

or say that life is where there are architectures,

any architectures, and we should stop fixating

on what is building the architectures to start with.

And the fact that the universe has discrete things

and it is completely mind blowing.

If you think about it for one second,

the fact there’s any objects at all,

and there’s, because for me, the object is a proxy

for a machine that built it,

some information being moved around,

actuation, sensing, getting resource,

and building these objects.

So for me, everyone’s been obsessing about the machine,

but I’m like, forget the machine, let’s see the objects.

And I think in a way that assembly theory,

we realized maybe a few months ago

that assembly theory actually does account for the soul

and the objects, not mystically like say,

Sheldrake’s morphic resonance,

or Leibniz’s monadology, seeing souls in things.

But when you see an object, and I’ve said this before,

but this object is evidence of thought,

and then there’s a lineage of those objects.

So I think what is fascinating is that,

you put it much more elegantly,

but the barrier between life and non life

is accruing enough memories to then actuate.

So what that means is there are contingency,

there are things that happen in the universe get trapped,

these memories then have a causal effect on the future.

And then when you get those concentrated in a machine,

and you’re actually able in real time,

able to integrate the past, the present with the future,

and do stuff, that’s when you are most alive.

You being the machine.


Wait a minute, why is the object,

so one of the ways to define life,

that Sarah said, is simple machines

creating complex machines.

So there’s a million questions there.

So how the hell does a simple machine

create a complex machine?

By mutation.

So this is what we were talking about at the beginning,

is you have the minimum replicator, so a molecule.

So this is what I was trying to convince Sarah

of the mechanism get there years ago, I think,

but then you’ve been building on it and saying,

you have a molecule that can copy itself,

but then there has to be some variability,

otherwise it’s not gonna get more functional.

So you need to add bits on.

So you have a minimum molecule that can copy itself,

but then it can add bits on,

and that can be copied as well,

and those add ons can give you additional function,

to be able to acquire more stuff to exist.

So existence is weird,

but the fact that there is existence is why there is life,

and that’s why I realized a few days ago

that there must be, that’s why alien life

must be everywhere, because there is existence.

Is there a conservation of cheeky stuff happening?

So how can you keep injecting more complex things?

Doesn’t the machine that creates the object

need to be as or more powerful than the things it creates?

So how can you get complexity from simplicity?

So the way you get complexity from simplicity

is that you, I’m just making this up,

but this is kind of my notion,

and you have a large volume of stuff,

so you’re able to get seeds, if you like,

random cues from the environment.

So you just use those objects

to basically write on your tape, ones and zeros, whatever.

And that is necessarily rich, complex, okay?

But it has a low assembliness,

but even though it has a high assembly number,

we can talk about that.

But then when you start to then integrate

that all into a smaller volume, as over time,

and you become more autonomous,

you then make the transition.

I don’t know what you think about that.

I think the easiest way to think about it is actually,

which I know is a concept you hate, but I also hate,

which is entropy, but people are more familiar with entropy

than what we talk about in assembly theory.

And also the idea that, like, say physics as we know it

involves objects that don’t exist across time,

or as we would say, low memory objects.

So one of the key distinctions that is…

Low memory objects.

Yeah, so physics is all…

Physicists are low memory objects.

Low memory objects.

Physicists are creators of low memory objects

or manipulators of low memory objects.


It’s a very nice way of putting it.

Okay, sorry, go ahead.

Yeah, no, it’s okay.

Sorry to keep interrupting.

No, no, no, it’s fine.

I like it too, it’s very funny.

But I think it’s a good way of phrasing it

because I think this kind of idea we have

in assembly theory is that, you know,

physics as we know it has basically removed time

as being a physical observable of an object.

And the argument I would make is that

when you look at things like water bottles or us,

we’re actually things that exist

that have a large extent in time.

So we actually have a physical size and time,

and we measure that with something called

the assembly index in molecules,

but presumably everyone should have sort of a,

do you want to explain what assembly?

Yeah, let’s, you know what?

Let’s step back and start at the beginning.

What is assembly theory?

Lee sent me some slides.

There’s a big sexy paper coming out probably.

Maybe, I don’t know.

We’ve almost finished it.

Almost, almost finished it.

That’s also a summary of science.

We’re almost done.


Well, no, no, we’re almost done.

It’s the history of science.

We are ready to start an interesting discussion

with our peers.


You’re the machine that created the object,

and we’ll see what the object takes us.

All right, so what is assembly theory?

Yeah, well, I think the easiest way

for people to understand it is to think about

assembly and molecules,

although the theory is very general.

It doesn’t just apply to molecules.

And this was really Lee’s insight,

so it’s kind of funny that I’m explaining it, but.

I’ll mark you.

Okay, all right, I’m ready, I’m ready.

You have to tell me where I get the check marks minus,


It’s your theory as well.

Yeah, I know.

But imagine a molecule,

and then you can break the molecule apart

into elementary building blocks.

They happen to be bonds.

And then you can think of all the ways,

for molecular assembly theory,

you can think of all the ways

of building up the original molecule.

So there’s all these paths that you can assemble it.

And the sort of rules of assembly is

you can use pieces that have been generated already.

So it has this kind of recursive property to it.

And so that’s where kind of memory

comes into assembly theory.

And then the assembly index is

the shortest path in that space.

So it’s supposed to be the minimal amount of history

that the universe has to undergo

in order to assemble that particular object.

And the reason that this is significant is

we figured out how to measure that

with a mass spec in the lab.

And we had this conjecture

that if that minimal number of steps

was sufficiently large,

it would indicate that you required a machine

or a system that had information

about how to assemble that specific object

because the combinatorial space of possibilities

is getting exponentially large

as the assembly index is increasing.

So it’s just, sorry to interrupt,

so that means there’s a sufficiently high assembly index

that if observed in an object

is an indicator that something lifelike created it

or is the object itself lifelike?


But you might want to make the distinction

that a water bottle is not life,

but it would still be a signature

that you were in that domain of physics

and that I might be alive.

So there will be potentially a lot of arguments

about where the line, at which assembly index

does interesting stuff start to happen.

The point is we can make all the arguments,

but it should be experimentally observable

and Lee can talk more about that part of it.

But the point I want to make about it is

there was always this intuition that I had

that there should be some complexity threshold

in the universe above which you would start to say

whatever physics governs life actually becomes operative.

And I think about it a little bit

like we have Planck’s constant,

and we have the fine structure constant.

And then this sort of assembly threshold

is basically another sort of potentially constant of nature.

It might depend on the specific features of the system,

but which we debate about sometimes.

But then when you’re past that,

you have to have some other explanation

than the current explanations we have in physics,

because now you’re in high memory.

The things actually require time for them to exist

and time becomes a physical variable.

The path to the creation of the object is the memory.

So you need to consider that.

Yeah, but the point is that’s a feature of the object.

So when I think of all the things in this room,

we see the projection of them as a water bottle,

but assembly theory would say that this is a causal graph

of all the ways the universe can create this thing.

That’s what it is as an object.

And we’re all interacting, a causal graph.

And most of the creativity in the biosphere

is because a lot of the objects that exist now

are huge in their structure across time.

Four billion years of evolution to get to us.

Is it possible to look at me

and infer the history that led to me?

If you, you as an individual might be hard.

You as a representative of a population of objects

that have high assembly

with similar causal history and structure

that you can communicate with, i.e. other humans,

you can infer a lot probably.

Yeah, also with them.

Which we do genomically even.

I mean, it’s not like,

we have a lot of information in us,

we can reconstruct histories from.

Assembly is saying something slightly deeper.

Yeah, one thing to add.

I mean, it’s not just about the object,

but the objects occur

and not just objects with high assembly number,

because you can have random things

that have a high assembly number,

but they must have,

there must be a number of identical copies.

So you know you’re getting away from the random,

because you could take a snapshot.

This is why, it’s not I hate entropy,

I love entropy when used correctly,

but it’s about the problem of entropy,

you have to have a labeler.

And so you can label the beginning and the end,

the start and the finish, you know,

where what you can do in assembly is say,

oh, I have a number of objects in abundance.

They all have these features.

And then you can infer.

And one of the things that we debated a lot,

particularly during lockdown,

because I almost went insane trying to crush the,

produce the assembly equation.

So we came up with the assembly equation.

I had, just imagine this.

So you have a string where,

actually it makes me sick trying to remember it.

It was so, it did my head in for a long time.

Yeah, because I couldn’t,

so if you just have a string of say words,

say, you know, a series of words, series of letters.

So you just have A, A, A, B, B, B, C, C, C, D, D, D.

And you find that object and you just have four A’s,

four B’s, four C’s, four D’s together, boom.

Then, and that you measured that.

So you physically measured that string of letters.

Then what you could do is you can infer sub graphs

of maybe the four A’s, the four B’s,

the four C’s and the four C’s,

but you don’t see them in the real world.

You just infer them.

And I really got stuck with that

because there’s a problem to try and work out

what’s the difference between a long,

you know, a physical object

and this assembly space of the objects

that we realized the best way to put that is infer in time.

So although we can’t infer your entire history,

we know at some point the four A’s were made,

the four B’s was made, the four C’s were made,

the four D’s were made and they all got added together.

And that’s one really interesting thing

that’s come out of the theory,

but the killer when we knew we were going beyond

beyond standard complexity theories,

but incredibly successful is that we realized

we could start to measure these things

for real across domains.

So the assembly index is actually an intrinsic property

of all stuff that you can break into components,

particularly molecules are good

because you can break them up

into smaller molecules, into atoms.

The challenge will be making that more general

across all the domains,

but we’re working on it right now

and I think the theory will do that.

So components, domains,

so you’re talking about basically measuring

the complexity of an object in what,

biology, chemistry, physics,

that’s what you mean by domains.

Complexity of tests.

Sociology. Complexity of computers.

Complexity of memes, you know.

Memes? Yep.

What is that, ideas?

Yeah, I mean, so one of the.

Ideas are objects in assembly theory.

Yeah. They are.

They’re physical things.

They’re just features of the causal graph.

I mean, the fact that I can talk to you right now

is because we’re exchanging structure

of our assembly space.

So conversation is the exchanging structures

in assembly space.

What is assembly space?

When I started working on origins of life,

I was writing about something called top down causation,

which a lot of like philosophers are interested in

and people that worry about the mind body problem.

But the whole idea is, you know,

if we have, you know, the microscopic world of physics

is causally complete,

it seems like there’s no room for higher level causes

like our thoughts to actually have any impact on the world.

And that didn’t, that seems problematic

when you get to studying life and mind

because it does seem that quote unquote,

emergent properties do matter to matter.

And then there’s this other sort of paradoxical situation

where information looks like it’s disembodied.

So we talk about information,

like it can just move from any physical system

to any other physical system.

And it doesn’t require,

like you don’t have to specify anything about the substrate

to talk about information.

And then there’s also the way we talk about mathematics

is also disembodied, right?

Like the platonic world of forms

and I think all of those things are hinging

that we really don’t know how to think about abstractions

as physical things.

And really, I think what assembly theory is pointing to

is what we’re missing there is the dimension of time.

And if you actually look at an object

being extended across time,

what we call information and the things that look abstract

are things that are entangled

in the histories of those objects.

They’re features of the overlapping assembly space.

So they look abstract because they’re not

part of the current structure,

but they’re part of the structure

if you thought about it as like the philosophical concept

of a hyperobject, an object that’s too big in time

for us to actually to resolve.

And so I think information is physical.

It’s just physical in time, not in space.

Too hyperobject, too difficult for us to resolve.

So we’re supposed to think about of life

as this thing that stretches through time

and there’s a causation chain that led to that thing.

And then you’re trying to measure something

with the assembly index about properties of that.

The assembly index is the ordering,

like you could think of it as like a partial ordering

of all the things that can happen.

So in thermodynamics, we coarse grain things

by temperature and pressure.

In assembly theory, we coarse grain

by the number of copies of an object

and the assembly index, which is basically,

if you think of the space of all possible things,

it’s like a depth of how far you’ve gone into that space

and how much time was required to get there.

In the shortest possible version.

The shortest possible version.

Not average, because can’t you just 3D?

You’re gonna kill me with that question.

Not 3D, can’t you always 3D print the thing?

Let’s like stab him in the heart.

No, because I had such a fight.

So Sarah’s team and my team are writing this paper

at the moment and.

It’s so funny.

I think we kind of share the, at the beginning,

you were like, no, that’s not right.

Oh yeah, that’s right.

And we’re doing this for a bit.

And then the problem is when you build a theory

and build the intuition,

there’s some certain features, right,

of the theory that almost felt like

I was being religious about saying,

right, you have to do this.

A good assembly theorist does this, does this, does this.

And Sarah’s postdoc, Daniel, and my postdoc, Abhishek,

and they were both.

They’re both brilliant.

They’re brilliant, but they were like,

no, we don’t buy that.

And I was like, it is, they were like,

well, Lee, actually, I thought you were the first

to say that, you know, you can’t,

if you can’t explain it, it doesn’t,

and you can’t do an experiment that doesn’t exist.

And that saved me.

And I said to Abhishek,

Abhishek’s my postdoc in Glasgow,

Daniel is Sarah’s postdoc in ASU.

I was like, I have the experimental data.

So when I basically take the molecules

and chop them up in the mass spec,

the assembly number is never the average.

It’s always the shortest.

It’s an intrinsic property.

And then the penny drop for Abhishek said, okay.

So I had these things that we had to believe

to start with or to trust,

and then we’d done the math and it comes out.

And they now have the shortest path, actually.

It’s up, it explains why the shortest path.

Here’s why the shortest path is important, not the average.

The shortest path needs you to identify

when the universe has basically got a memory,

not an average.

So what you want to be able to do is to say,

what is the minimum number of features

that I want to be able to see in the universe?

When I find those features,

I know the universe has had a coherent memory

and is basically alive.

And so that gives you the lower bound.

So that’s like, of course there’s going to be other paths.

We can be more ridiculous, right?

We can have other parts, but it’s just the minimum.

So probabilistically at the beginning,

because assembly theory was built

as a measure for biosignatures, I needed to go there.

And then I realized it was intrinsic.

And then Sarah realized it was intrinsic

and these hyperobjects were coming.

And we were kind of fusing that notions together.

And then the team were like, yeah,

but if I have enough energy and I have enough resources,

I might not take the shortest path.

I might go a bit longer.

I might take a really long path

because it allows me then to do something else.

So what you can do is, let’s say

I’ve got two different objects, A and B,

and they both have different shortest paths to get them.

But then if you want to make A and B together,

they will have a compromise.

So in the joint assembly space, that might be an average,

but actually it’s the shortest way

you can make both A and B

with a minimum amount of resource in time.

So suddenly you then layer these things up.

And so the average becomes not important,

but as you literally overlap those sets,

you get a new shortest path.

And so what we realized time and time again

when we’re doing the math,

the shortest path is intrinsic, is fundamental,

and is measurable, which is kind of mind blowing.

So what we’re talking about, some basic ingredients,

maybe we’ll talk about that, what those basic ingredients

could be and how many steps, when you say shortest path,

how many steps it takes to turn those basic ingredients

into the final meal.

So how to make a, what’s the shortest way to make a pizza?

Or a pie.

An apple pie.

That’s right.

And a pizza and a pie together.

Or a scratch.

So there’s a lot of ways.

There’s the shortest way,

and then you take the full spectrum of ways

and there’s probably an average duration

for a noob to make an apple pie.

Is the average interesting still?

If you measure the average length of the path

to assemble a thing, does that tell you something

about the way nature usually does it?

Versus something fundamental about the object,

which I think is what you’re aiming at

with the assembly index.

Yeah, I mean, look, we all have to quantify things.

The minimum path gives you the lower bounds.

You know you’re detecting something.

You know you’re inferring something.

The average tells you about really how the objects

are existing in the ecosystem or the technology.

And there has to be more paths explored

because then you can happen upon other memories

and then condense them down.

I’m not making too much sense, but if you look and say,

let’s just say, I mean, maybe we’re gonna get

to alien civilizations later, right?

But I would argue very strongly

that alien civilization A and alien civilization B,

they’re different assembly spaces.

So they’re kind of gonna be a bit messed up

if they happen to come one another,

only when they find some joint overlap in their technology,

because if aliens come to us and they don’t share

any of the causal graph we’ve showed,

but hopefully they share the periodic table

and bonds and things,

that we’re gonna have to really think about the language

to talk to us aliens by inferring,

by using assembly theory to infer their language,

their technology, and other bits and bobs.

And the shortest path will help you do that quickly.

All right, so all aliens in this causality graphs

have a common ancestor in the…

If the building blocks are the same,

which means they live in the same universe as us.

So this is the assumption.

It depends on how far back in time you go, though.

But the universe has all the same building blocks.


And we have to assume that.

So at least there’s not different classes

of causality graphs, right?


The universe doesn’t just say like,

here you get the red causality graph,

and you get the blue one.

These basic ingredients,

and they’re geographically constrained,

or constrained in space or time, or something like that.

They’re constrained in time

because only by the virtue of the fact

that you need enough time to have passed

for some things to exist.

So the universe has to be big enough in time

for some things.

So just one point on the shortest path

versus the average path,

which I think we’ll get to this,

is you had a nice way of saying it’s like

the minimal compression is the shortest path

for the universe to produce that.

But it’s also like the first time

in the ordering of events

that you might expect to see that object.

But the average path tells you something

about the actual steps that were realized,

and that becomes an emergent property

of that object’s interaction with other objects.

So it’s not an intrinsic feature of that object.

It’s a feature of the interactions with other things.

And so one of the nice features of assembly

is you’ve basically gotten rid of,

you just look at the things that exist,

and you’ve gotten rid of the mechanisms

for constructing them in some sense.

Like the machines are not as important

in the current construction of the theory,

although I would like to bridge it

to some ideas about constructors.

But then you can only communicate with things

as Lee was saying,

if you have some overlap in the past history.

So if you had an alien species

that had absolutely no overlap,

then there would be no means of communication.

But as we progress further and further in time

and more things become possible

because the assembly spaces are larger,

because you can have a larger assembly space

in terms of index and also just the size of the space,

because it’s exponentially growing,

then more things can happen in the future.

And the example I like to give is actually

when we made first contact with gravitational waves,

because that’s an alien phenomenon

that’s been permeating our planet,

not alien in life phenomenon,

but alien like something we had never knew existed.

It’s been like there’s gravitational waves

rippling through this room right now,

but we had to advance to the level of Einstein

writing down his theory of relativity

and then 100 years of technological development

to even quote unquote see that phenomena.

So the, okay, to see that phenomena,

our causal graph have to start intersecting.

Yeah, we needed the idea to emerge first,

the abstraction, right?

And then we had to build the technology

that could actually observe features of that abstraction.

So the nice promising thing is over time,

the graph can grow so we can start overlapping eventually.

Yeah, so the interesting feature of that graph

is there was an event 1.4 billion years away

of a black hole merger that we detected on our detector.

And now suddenly we’re connected

through this communication channel

with this distant event in our universe

that if you think about 1.4 billion years ago,

what was happening on this planet

or even further back in time,

that there’s common physics underlying all of those events,

but even for those two events to communicate.

Now I understand what you were going on about

the other week.

Yeah, I’m sorry, this is a really abstract example,

but it’s sort of.

Your causal graphs are now overlapping.

Well, let’s just say now our causal graphs

are overlapping in the deep past.

Yeah, sorry.

No, I like it, so you made it.

I totally missed it.

Oh, the 1.6 billion.


You made a connection with it.

No, I do like that.

No, no, you can tell me what your epiphany is now.

That’s good.

Because I was.

And I should get the jokes before 30 seconds after, so.

Oh, I get it now.

No, it’s all right.

I was slow.

The joke from two minutes ago.

I’m slow on the uptake here.

I wasn’t able to comprehend

what you were talking about when saying

the channel communicating to the past.

But what you’re saying is we were able to infer

what happened 1.4 billion years ago.

We detected the gravity wave.

I mean, I think it’s amazing that at that time,

we weren’t even, we were just becoming multicellular,


It’s like insane.

And then we progressed from multicellularity

through to technology and built the detector

and then we just extrapolate backwards.

So, although we didn’t do anything back to the graph

back in time, we understood this existence

then overlapped going forward.

And that.

Well, that’s because our graphs are larger.

Yeah, but that means that has a consequence.

One of the things I was trying to say is I think,

I don’t know, Sarah might be, she can correct me,

information first and I’m a object first kind of guy.

So, I mean, there’s things that get constructed.

There has to be this transition in random constructions.

So when the object that’s being constructed by the process

bakes in that memory and those memories then add on

and add on and add on.

So as it becomes more competent and life is about

taking those memories and compressing them,

increasing their autonomy.

And so I think that, like the cell that we have

in biology on earth is our way of doing that,

that really the maximum ability to take memories

and to act on the future.

Oh, I think that’s mathematics.

No, mathematics doesn’t exist.

No, but that’s the point.

The point is that abstractions do exist.

They’re real physical things.

We call them abstractions.

But the point about mathematics that I think is,

so I don’t disagree that I think you’re object first

and I’m information first, but I think I’m only

information first in the sense that I think the thing

that we need to explain is what abstractions are

and what they are as physical things

because of all of human history,

we’ve thought that there were these properties

that are disembodied, exist outside of the universe

and really they do exist in the universe

and we just don’t understand what their physics is.

So I think mathematics is a really good example.

We do theoretical physics with math,

but imagine doing physics of math

and then thinking about math as a physical object.

And math is super interesting.

I think this is why we think it describes reality so well

because it’s the most copyable kind of information.

It retains its properties

when you move it between physical media,

which means that it’s very deep.

And so it seems to describe the universe really well,

but it probably is because it’s information

that’s very deep in our past.

And it’s just, we invented a way of communicating it

very effectively between us.

Isn’t math more fundamental?

Isn’t the assembly of the graph,

isn’t basically, I sound completely boring.

It’s like math, assembly theory invented math, but it did.

It has to be.


So what is math exactly?

It’s a nice simplification,

a simple description of what?

So we have a computer scientist,

a physicist, and a chemist here.

Walking to a bar.

I think the chemist is gonna define math

and you guys can correct me.

Go for it.

I would say.

Lay in on us, Lee.

We’re ready.

I think the ability to label objects

and place them into classes

and then do operations on the objects is what math is.

So on that point,

what does it mean to be object first

versus information first?

So what’s the difference between object and information

when you get to that low fundamental level?

Well, I might change my view.

So I’m stuff first, the stuff.

And then when stuff becomes objects,

it has to invent information.

And then the information acts on more stuff

and becomes more objects.

So I think there is a transition to information

that occurs when you go from stuff to objects.

You mean time though, I think.


Information is emergent.

Not emergent.

Information is actionable memories from the universe.

So when memories become actionable, that’s information.

But there’s always memory, but it’s not actionable.


And then it’s not information.


And actionable is what you can create.

You can use it.

If you can’t use it, then it’s not information.

If you can’t transmit it,

if it doesn’t have any causal consequence.

Falls in the forest.

I don’t understand.

Why is that not information?

It’s not information.

It’s stuff.

It’s stuff happening, but it’s not causal.

Yeah, yeah, we can.

This is cool.

But it’s happening.

Happening requires information.

No, no, no, no, no.

Stuff is always happening.

No, this is where the physicists

and the math petitions get themselves in a loop

because I think the universe, I mean,

I think say Max Tegmark is very playful

and say like the universe is just math.

Well, the universe is just math.

Then we might as well not bother having any conversation

because the conversation already written,

we just might as well go to the future and say,

can you just give us the conversations happened already?

So I think the problem is that math petitions

are so successful at labeling stuff

and so successful understanding the stuff

through those labels,

they forget that actually those labels had to emerge

and that information had to be built on those memories.

So memory in the universe, so constraints, graph,

when they become actionable

and the graph can loop back on itself

or interact with other graphs and they can intersect,

those memories become actionable

and therefore they’re information.

And I think you just changed my mind

on something pretty big, but I don’t have a pen.

So I can’t write, I’m gonna write it down later,

but roughly the idea is like you’ve got these two graphs

of objects of stuff that you have memories

and then when they intersect

and then they can act on each other,

that’s maybe the mechanism by which information is then,

so then you can then abstract.

So when one graph can then build another graph and say,

hey, you don’t have to go through the nonsense

we had to go through.

Here’s literally the way to do it.

Stuff always comes first,

but then when stuff builds the abstraction,

the abstractions can be then teleported

onto other stuff.

And the abstractions is the looping back power.


Am I making, I don’t know, I got stuck.

Yeah, so first, God made stuff.

Then after that, when you start to be able

to form abstractions, that’s when the information.

God is the memory the universe can remember.

Otherwise, there’s no, wait, did you?

Someone’s gonna be deciphering that statement

hundreds of years from now, what the hell does that mean?

What does the humans mean by this?

Look, don’t diss my one liners.

I took me 15 seconds to come up with that.

I don’t know what it means.

What does it mean?

Okay, wait, we need to, how do we get onto this?

We were time, causality, mathematics.

So what is mathematics in this picture of stuff,

objects, memory, and information?

What exactly is mathematics?

It’s the most efficient labeling scheme

that you can apply to lots of different graphs.

Labeling scheme doesn’t make it sound useful.

Can I try?

Yep, sure, please.

Have you rejected my definition of mathematics?

I’m shocked.

Yeah, no, it’s all right.

But it’s correct.

Go on, sorry.


No, I mean, I think we have a problem, right?

Cause we can’t not be us,

like we’re stuck in the shells we are

and we’re trying to observe the world.

And so mathematics looks like it has certain properties.

And I guess the thought experiment I find is useful

is to try to imagine if you were outside of us

looking at us as physical systems using mathematics,

what would be the specific features you associate

to the property of understanding mathematics

and being able to implement it in the universe, right?

And when you do that,

mathematics seems to have some really interesting properties

relative to other kinds of abstraction

we might talk about like language or artistic expression.

One of those properties is the one I mentioned already

that is really easy to copy between physical media.

So if I give you a mathematical statement,

you almost immediately know what I mean.

If I tell you the sky is blue,

you might say, is it gold ball blue?

Is it azure blue?

What color blue do you mean?

And you have a harder time visualizing what I actually mean.

So mathematics carries a lot of meaning with it

when it’s copied between physical systems.

It’s also the reason we use it to communicate with computers.

And then the second one is it retains its property

of actually what it can do in the universe when it’s copied.

So the example I like to give there

is think about like Newton’s law of gravitation.

It’s actually, it’s a compressed regularity

of a bunch of phenomena that we observe in the universe,

but then that information actually is causal in a sense

that it allows us to do things we wouldn’t be able to do

without that particular knowledge

and that particular abstraction.

And in this case, like launch satellites to space

or send people to Mars or whatever it is.

So if you look at us from the outside and you say,

what is it for physical systems

to invent a thing called mathematics

and then to use and then it to become a physical observable,

mathematics is kind of like

the universally copyable information

that allows new possibility spaces

to be open in the future

because it allows this kind of ability

to map one physical system to another

and actually understand that the general principles.

So is it helping the overlap of causal graphs then

by mapping?

Oh, I think that’s the explanation for what it is

in terms of the physical theory of assembly

would be some feature of the structure

of the assembly spaces of causal graphs

and their relationship to each other.

So for example, and I mean, this is things

that we’re gonna have to work out over the next few years.

I mean, we’re in totally uncharted conceptual territory

here, but as is usual, diving off the deep end,

but I would expect that we would be able to come up

with a theory of like, why is it that some physical systems

can communicate with each other?

Like language, language is basically

because we’re objects extended over time

and some of the history of that assembly space

actually overlaps.

And when we communicate,

it’s because we actually have shared structure

in our causal history.

Let me have another quick go at this, right?

So I think we all agree.

So I think we take mathematics for granted

because we’ve gone through this chain, right?

Of, you know, we all share a language now, okay?

And we can, well, we share language,

so we have languages that we can make interoperable.

And so whether you’re speaking, I don’t know,

all the different dialects of Chinese,

all the different dialects of English, French, German,

whatever, you can interconvert them.

The interesting thing about mathematics now

is that everybody on planet Earth, every human being

and computers share that common language.

That language was constructed by a process in time.

So what I’m trying to say is assembly invented math

is those, right from the, you know,

mathematics didn’t occur, it didn’t exist before life.

Abstraction was invented by life, right?

That doesn’t mean that the universe wasn’t capable

of mathematical things.

Wait a minute, can we just ask that old famous question,

is math invented or discovered?

So when you say assembly invented, or whatever.

It means it’s just.

Well, someone might just say assembly

is a mathematical theory, but sorry.


Are we arguing? Exactly.

Are we arguing now?

That’s what it sounds like.

Are we discovering mathematics?

No, well, yes and no.

I would say.

And you call mathematics a language

that we’re developing. I would say that,

look, I’m pretty sure that there are some very common

seeds of mathematics in the universe, right?

But actually not the mathematics that we are finding now

is not discovered, it’s invented.

And, but even though I think those two terms

are very triggering, and I don’t think

they’re necessarily useful,

because I think that what people do,

the mathematicians that say, oh, mathematics was discovered

because they live in a universe where there is no time

and it just all exists.

But what I’m saying is, and I think in the same way

you can create, let’s say I’m gonna go and create

and make a piece of art, did I make that piece of art

or did I discover it?

Like inventing the aeroplane.

Did I invent the aeroplane?

Let’s stick with the aeroplane.

The aeroplane is a good one.

Let’s say, did I discover the aeroplane?

Well, in a way, the universe discovered the aeroplane

because it’s just chucked a load of atoms together

and a load of random human beings want to do stuff

and then we discover the aeroplane

in the space of all the possibilities.

But here’s the thing, when the space of possibilities

is so vast, infinite almost, and you’re able to actualize

one of those in an object, then you are inventing it.

So in mathematics, because there are infinite number

of theorems, the fact you’re actually pulling,

there’s no difference between inventing

a mathematical structure and inventing the aeroplane.

They’re the same thing, but that doesn’t mean

that now the aeroplane exists in the universe,

it’s something weird about the universe.

So I think that the more, this is the thing

that you probably, the more memory required

for the object, the more invented it is.

So when a mathematical theorem needs more bytes

to store it, the more invented it is,

and the less bytes, the more discovered it is.

But everything then is invented.

It’s just more or less invented.



The universe has to generate everything as it goes.


And it wasn’t there in the beginning.

And the way we’re thinking it,

when you’re thinking about the difference

between invented and discovered,

is because we’re throwing away all the memory.


So if you start to think in terms of causality and time,

then those things become the same.

Everything is invented.

And the idea is to make everything intrinsic

to the universe.

So I think one of the features of assembly theory

is we don’t wanna have external observers.

There’s been this long tradition in physics

of trying to describe the universe from the outside

and not the inside.

And the universe has to generate everything itself

if you do it from the inside.

Assembly theory describes how the universe builds itself.

Did it take you 15 seconds to say that?


And to come up with that also?

No, I’ve thought of that before.


That’s a good line.

It’s a, it’s like.

Oh, you’re making fun of me.

No, I’m not making fun, I’m having fun.

There’s a difference.

Oh, that’s good.

All right.

She’s inventing fun.

I’m not all intimidated.


And there’s a causal history to that fun.

You mentioned that there’s no way to communicate

with aliens until there’s overlap in the causal graph.

Communication includes being able to see them.

And like, what are we, this is the question is,

is communication any kind of detection?

And if so, what do aliens look like

as you get more and more overlap on the causal graph?

You’re assuming, let’s assume that aliens,

so when you see them and they see you,

you’re assuming they have vision,

they have the ability to construct in 3D and in time.

That’s a lot of assumptions they’re making.

What detection?

All right, let’s step back.

So yes, okay, you’re right.

So when, in the English language,

when we say the word see, we mean visually,

they show up to a party and it’s like,

oh, wow, that’s an alien.

That’s visual, that’s 3D, that’s, okay.

And that’s also assuming scale,

spatial scale of something that’s visible to you.

So it can’t be microscopic or it can’t be so big

that you don’t even realize that’s an entity.

Okay, but other kinds of detection too.

I would make it more abstract and go down.

I was thinking this morning about how to rewrite

the Arecibo message in assembly theory

and also to abandon binary.

Because I don’t think aliens necessarily,

why should they have binary?

Well, they have some basic elements

with which to do information exchange.

Let’s make it more fundamental, more universal.

So we need to think about what is the universal way

of making a memory and then we should

reencode Arecibo in that way.

What’s more basic than zeros and ones?

Well, it’s really difficult to get out

that causal chain because we’re so,

so let’s erase the idea of zero for a moment.

It took human beings a long time

to come up with the idea of zero.

Now you got the idea of zero, you can’t throw it away.

It’s so useful.

To discover the idea of zero.

To discover or invent.

I don’t know, but it took a long time

so it was invented, that’s right.

Yeah, I think zero was invented, exactly.

So it’s not a given that aliens know what zero is.

That’s a massive assumption.

It’s a useful discovery.

You’re saying if you break the causal chain

there might be some other more efficient way of representing.

That’s why I wanna meet him and ask him.

For a shortcut.

But you won’t be able to ask him until.

So I interrupted you and I think you’re making good point.

I was just gonna say, well look.

Thank you.


Rather than saying.

Please internet, tweet at him for the rude interruptions.

Oh, go ahead, I’m sorry.

No, it’s okay.

Maybe it’s change.

How do we, so, oh, I don’t know what it’s like

to be an alien.

I would like to know.

What is the full spectrum of what aliens

might look like to us?

Now that we’ve laid this all on the table of like,

all right, so there has to be some overlap

in this causal chain that led to them.

What are we looking for?

What do you think we should be looking for?

So you mentioned mass spec.

Measuring certain objects that aliens could create

or are aliens themselves.

We show up to a planet or maybe not a planet.

Or maybe, what the hell is the basic object

we’re trying to measure the assembly index of?

Let’s cut ourselves a break.

Let’s assume that they are metabolized.

They’ve got an energy source.

And they’re a size that we can recognize.

Let’s give ourselves a break.

Because there could be aliens that are so big

we won’t recognize we’re seeing them.

There might be aliens that are so small

we don’t yet have the ability to,

we don’t have microscopes that can see far enough away

that just wouldn’t be able to see them.

So what’s a good range?

So let’s just make a range.

Let’s just be very anthropocentric and say,

we’re gonna look for aliens roughly our size

and technology our size.

Because we know it’s possible on Earth, right?

I mean, a reasonable thing to do would be

to find exoplanets that are in the same zone as Earth

in terms of heat and stuff.

And then say, hey, if there’s that same kind of gravity,

same kind of stuff, we could reasonably assume

that alien life there might use a similar kind

of physical infrastructure.

And then we’re good.

So then your question becomes really relevant.

Say, right, let’s use vision, sound, touch.

And then.

So okay, that’s really nice.

So that if there’s a lot of aliens out there,

there’s a good likelihood if you match to the planet

that they’re going to be in the same spatial

and temporal, operating in the same spatial

and temporal domain as humans.

Okay, within that, what do they look like visually?

What do they sound like?

What do they, oh god, this sounds creepy.

Tastes like, what do they, oh, smell like, smell like.

That sounds like our clubhouse.

We was like, can we have sex with aliens?

Which was basically me saying.

Passionate, passionate love.

But it wasn’t actually about sex.

It was about, is our chemistry compatible, right?

Is there some?


Yeah, can we, yeah.

Are they edible too?

They could be very edible.

They could be delicious.

That’s why I want to see some aliens, right?

Because I think, are there, I think evolution,

I mean, evolution exploits symmetry, right?

Because why generate memory?

Why generate storage, the need for storage space

when you can use symmetry?

So, and symmetry is quite, may be quite effective

in allowing you to mechanically design stuff, right?

So maybe alien, you could be reasonable to assume

that aliens could have, they could be bipedal.

They could be symmetric in the same way.

Might have a couple of eyes or a couple of senses.

We can make them, perhaps there’s this whole zoo

of different aliens out there.

And we’ll never get to be able to classify

some of the weird aliens we can’t interact with

because they have made such weird stuff.

But we are just going to look at,

we’re going to find aliens that look most like us.

Why not?

Because those are the first ones we’re likely to see.


But I think it’s really hard to imagine

what the space of aliens is because the space is huge.

Because, you know, like one of the arguments

that you can make about why life emerges in chemistry

is because chemistry is the first scale

in terms of like, you know,

building up objects from elementary objects.

That the number of possible things that could exist

is larger than the universe can possibly make all at once.


So imagine you have two planets

and they’re cooking some geochemistry.

You know, our planet invented one kind of biochemistry.

And presumably as you start building up

the complexity of the molecules,

the chances of the overlap in those trajectories,

those causal chains being built up is probably very low.

And it gets lower and lower as it gets further advanced

along its evolutionary path.

So I think it’s very difficult to imagine

predicting the technologies that aliens are gonna have.

I mean, it’s so, you’re looking at basically

planets have kind of convergent chemistry,

but there’s some variability.

And then you’re looking basically at the outgrowth

into the possibility space for chemistry.

So do you think we would detect the technology,

the objects created by aliens before we detect the aliens?


So when you’re talking about measuring assembly index,

don’t you think we would detect the garbage first?

Like at the outskirts of alien civilizations,

isn’t this just gonna be trash?

I think I would come back to Arecibo.

The Arecibo message sent from the Arecibo telescope

built by Drake, I think, and Sagan.

How’s Arecibo spelled?

A R E C I B O.

Yes, thank you.

And there we go, they’ve got that up there.

That’s the telescope that sent the message

that you’re talking about.

So that message was sent where?

It was beamed at a star, a specific star,

and it was sent out many years ago.

And what they did, so this is why I was pushing on binary,

it’s a binary message.

I think it’s a semi prime length number of characters.

So I think 73 by 23, I think.

And it basically represents human bit proton,

binary, human beings, DNA, male and female.

And it’s really cool.

But I’m just wondering if it could be done

not making any,

cause it made assumptions that aliens speak binary.

Why make that assumption?

Why not just assume that if the difference between physics,

chemistry and biology is the amount of memory

that’s instead that’s recordable by the substrates,

then surely the universal thing,

I’m gonna make some sacrilegious statement,

which I think is pretty awesome for people to argue with.

So this is, we’re looking at an image

where it’s the entirety of the message encoded in binary.

And then there’s a probably interpretation

of different parts of that image.

There’s a person, there’s green parts.

It looks like for people just listening,

like a game of Tetris.

So it’s encoding in minimal ways,

a bunch of cool information, probably.

Representing all of us.

So the topic’s kind of teaching us how to count

and then it all goes all the way down

teaching you chemistry and then just says,

but it makes so many assumptions.

And I think if we can actually,

so look, I think, I mean, Sarah’s much more eloquent

in expressing this, but I’ll have a go

and you can correct it if you want,

which is like one of the things that Sarah has had

a profound effect on the way I look at the origin of life.

And this is one of the reasons why we’re working together

because we don’t really care about the origin of life.

We wanna make life, make aliens and find aliens.

Make aliens, find aliens.

I think we might have to make aliens in the lab

before we find aliens in the universe, right?

I think that would be a cool way to do it.

So what is it about the universe that creates aliens?

Well, it’s selection through assembly theory,

creating memories, because when you create memories,

you can then command your domain.

You can basically do stuff.

You can command matter.

So we need to find a way by understanding what life is

of how the minimal way to command matter,

how that would emerge in the universe.

And if we want to communicate,

I mean, maybe we don’t want

to necessarily uniformly communicate.

What I would do perhaps if I had,

is I would send out lots of probes away from Earth

that have this magic way of communicating with aliens,

get them quite a far away from Earth, plausibly deniable,

and then send out the message

that would then attract all the aliens,

and then basically work out if they were friend or foe

and how they wanna hang out.

The messages being something has to do with the memories.

Yes, like the assembly version of Arecibo,

so that everyone in the universe

that has been understands what life is.

So aliens need to work out what they are.

Once they’ve worked out what they are,

they then can work out how to encode what they are,

and then they can go out and send messages.

It’s like the universal, the Rosetta Stone

for life in the universe is working out

how the memories are built.

I don’t know, Sarah, you have any, well,

whether you would agree with that.

No, I wanted to raise a different point,

which is about the fact that we can’t see the aliens yet

because we haven’t gotten the technology.

And presumably we think assembly theory

is the right way of doing it,

but I don’t think that we know how to go

from the kind of data you’re describing, Lex,

like visual data or smell

to construct the assembly spaces yet.

And in some ways, I think that the problem

of life detection really is the same problem

at the foundations of AI that we don’t understand

how to get machines to see causal graphs,

to see reality in terms of causation.

And so I think assembly and AI

are gonna intersect in interesting ways, hopefully,

but the sort of key point,

and I’ve been trying to make this argument more recently,

I might write an essay on it,

is people talk about the great filter, right?

And which is, again, this doomsday thing

that people wanna say there’s no aliens out there

because something terrible happened to them.

And it matters whether that’s in our past or our future

as to the longevity of our species, presumably,

which is why people find it interesting.

But I think it’s not a physical filter.

It’s not like things go extinct.

I think it’s literally,

we don’t have the technology to see them.

And you could see that with microscopes.

I mean, we didn’t know there were microbes on this table

or tables for thousands of years or telescopes.

Like there’s so much of the universe we can’t see.

And then basically what we have done as a species

is outsource our physical perceptions to technology,

building microscopes based on our eyes,

and building seismometers based on our sense of feelings,

like feel earthquakes and things.

And AI is basically we’re trying to outsource

what’s actually happening in our thinking apparatus

into machines now and to technological devices.

And maybe that’s the key technology

that’s gonna allow us to see things like us

and see the universe in a totally different way.

But you kind of mentioned the great filter.

Do you think there’s a way through technology

to stop being able to see stuff?

So can you take a step backwards?

I think so, yeah.

Did you imply that with the great, so like?

Well, no, I mean, I think there’s a great perceptual filter

in the sense that a example of life evolving on a planet

over billions of years has to acquire

a certain amount of knowledge and technology

to actually recognize the phenomena that it is.

Well, that’s the sense I have is,

I mean, you talk with physicists, engineers in general,

there’s this kind of idea that we have

most of the tools already to hear the signal.

But to me, it feels like we don’t have any of the tools

to see the signal.

No, we don’t know what we’re doing, yeah, I agree.

That’s the biggest, like to hear.

We don’t have the tools to really hear, to see.


Aliens are everywhere.

We just don’t have the, yeah, well, oh, that’s.

I mean, I got this in part, actually,

because you were like, you know,

last time I was here, he was like, look at the carpet.

You know, if you had an alien detector

where the carpet be aliens.

I mean, I think we really don’t.

I think it would be.

But the aliens would nevertheless have a high assembly index

or produce things of high assembly index.

And those things of high assembly index,

you have to have a detector that can recognize

high assembly index in all its forms.

Yeah. Yes.

That’s it, that’s it.

Take data, construct assembly space.

Yeah. Those patterns, basically.

So one way to think about high assembly index

is interesting patterns of basic ingredients.

I can give you an example,

because I mean, in molecules,

we’ve been talking about in objects,

but we’re also trying to do it in spatial trajectories.

Like, imagine you’re just,

like, I always get bothered by the fact that, like,

when you look at birds flocking,

you can describe that with like a simple Boyd’s model,

or like, you know, people use spin glass

to describe animal behavior.

And those are like really simple physics models.

Yet you’re looking at a system that you know has agency

and there’s intelligence in those birds.

And basically, like, you can’t help but think

there must be some statistical signatures

of the fact that they’re,

that’s a group of agents versus, you know, like,

I don’t know, you know, the physics example,

maybe like, I don’t know, Brownian motion or something.

And so what we’re trying to do

is actually apply assembly to trajectory data

to try to say there’s a minimal amount of causal history

to build up certain trajectories for observed agents

that’s like an agency detector for behavior.

Do you think it’s possible to do some like Boyd’s

or those kinds of things, like artificial,

like cellular automata, play with those ideas

with assembly, with assembly theory?

Have you found any useful, really simple mathematical,

like, simulation tools that allow you

to play with these concepts?

So like one, of course, you’re doing mass spec

in this physical space with chemistry,

but it just seems, well, I mean,

computer science person, maybe,

it seems easier to just.

I agree with you.

It seems even sexier in terms of tweeting visual information

on Twitter or Instagram, more importantly,

to play like, here’s an organism of a low assembly index

and here’s an organism of a high assembly index

and let’s watch them create more and more memories

and more and more complex objects.

And so like, in mathematics,

you get to observe what that looks like

to build up an intuition what assembly index is like.

We are building a toolkit right now.

So I think it’s a really good idea,

but what we’ve got to do is I’m kind of still obsessed

with the infrastructure required.

And one of the reasons why I was pushing on information

and mathematics when human beings,

when human beings, we take a lot of the infrastructure

for granted.

And I think we have to strip that back a bit

for going forward, but you’re absolutely right.

I would agree that I think the fact that we exist

in the universe, this is like,

I can see that lots of people would disagree

with the statement, but I don’t think Sarah will,

but I don’t know.

The fact that objects exist,

I don’t think anyone on earth will disagree

that objects can exist elsewhere, right?

But they will disagree that life can exist elsewhere.

But what perhaps I’m trying to say is that

the acquisition, the universe’s ability to acquire memory

is the very first step for building life.

And that must be, that’s so easy to happen.

So therefore alien life is everywhere

because all alien life is,

is those memories being compressed and minimalized

and the alien equivalent of the cell working.

So I think that we will build new technologies

to find aliens, but we need to understand what we are first

and how we go from physics to chemistry to biology.

The most interesting thing,

as you’re saying to these two organisms,

different assemblies, there’s one you get into biology.

Biology gets more and more weird,

more and more contingent.

Physics is, chemistry is less weird

cause the rules of chemistry are smaller

than the rules of biology.

And then going away to physics where you have a very

nicely tangible number of ways of arranging things.

And I think assembly theory just helps you appreciate that.

And so once we get there,

my dream is that we are just gonna be able to suddenly,

I mean, I’m maybe just being really arrogant here.

I don’t mean to be arrogant.

It’s just, I’ve got this hammer called assembly

and everything’s a nail.

But I think that once we crack it,

we’ll be able to use assembly theory plus telescopes

to find aliens.

Do you have, Sarah, do you have disagreements with Lee

on the number of aliens that are out there?

I do actually, yeah, well.

And what they look like.

So any of the things we’ve been talking about,

is there nuanced, it’s always nice to discover wisdom

through nuanced disagreement.

Yeah, I don’t wholly disagree, but I think,

but I do think I disagree.

It’s kind of, there’s nuance there.

But Lee made it.

You can disagree.

No, it’s fine.

It is nuanced, right?

So you made the point earlier that you think,

once we discover what life is,

we’ll see alien life everywhere.

And I think I agree on some levels in the sense

that I think the physics that governs us is universal.

But I don’t know how far I would go to say,

to say that we’re a likely phenomenon

because we don’t understand all of the features

of the transition at the origin of life,

which we would just say in assembly,

as you go from the no memory physics

to there’s like a critical transition

around the assembly index

where assembliness starts to increase.

And that’s what we call the evolution of the biosphere

and complexification of the biosphere.

So there’s a principle of increasing assembliness

where that goes back to what I was saying

at the very beginning about the physics of the possible,

that the universe basically gets in this mode

of trying to make as much possibilities as possible.

Now, how often that transition happens

that you get the kind of cascading effect

that we get in our biosphere, I think we don’t know.

If we did, we would know the likelihood of life

in the universe.

And a lot of people wanna say life is common,

but I don’t think that we can say that yet

till we have the empirical data,

which I think you would agree with.

But then there’s this other kind of thought experiment I have

which I don’t like, but I did have it,

which is if life emerges on one planet

and you get this real high density of things

that can exist on that planet,

is it sort of dominating the density of creation

that the universe can actually generate?

So like if you’re thinking about counting entropy, right?

Like the universe has a certain amount of stuff in it.

And then assembly is kind of like an entropic principle.

It’s not entropy.

But the idea is that now transformations among stuff

or the actual physical histories of things

now become things that you have to count

as far as saying that these things exist

and we’re increasing the number of things that exist.

And if you think about that cosmologically,

maybe Earth is sucking up all the life potential

of the whole universe, I don’t know.

But I haven’t.

How’s that, can you expand that a little bit?

Why can any one geographical region

suck up the creative capacity of the universe?

Just like, I know it’s a ridiculous thought.

I don’t actually agree with it,

but it was just a thought experiment.

I love that you can have thoughts

that you don’t like and don’t agree with,

but you have to think through them anyway.

The human mind is fascinating.

Yeah, I think these sort of counterfactual

thought experiments are really good

when you’re trying to build new theories

because you have to think through all the consequences.

And there are people that want to try to account for,

say, the degrees of freedom on our planet

in cosmological inventories of talking about

the entropy of the universe.

And when we’re thinking about cosmological

arrow of time and things like that.

Now, I think those are pretty superficial proposals

as they stand now, but assembly would give you

a way of counting it.

And then the question is if there’s a certain

maximal capacity of the universe’s speed

of generating stuff, which Lee always has this argument

that assembly is about time.

The universe is generating more states.

Really what it’s generating is more assembly possibilities.

And then dark energy might be one manifestation of that,

that the universe is accelerating its expansion

because that makes more physical space.

And what’s happening on our planet is it’s accelerating

in the expansion of possible things that exist.

And maybe the universe just has a maximal rate

of what it can do to generate things.

And then if there is a maximal rate,

maybe only a certain number of planets

can actually do that.

Or there’s a trade off about the pace of growth

on certain planets versus others.

I have a million questions there,

but do you have thoughts on?

Just a quick, yeah, I’ll just say something very quick.

It’s a thought experiment.

No, it’s good, I think I get it.

I think I get it.

So what I want to say is when I mean aliens are everywhere,

I mean memories are the prerequisite

for aliens via selection

and then concentration of selection

when selection becomes autonomous.

So what I would love to do is to build,

say a magical telescope that was a memory,

a magical one, or a real one,

that would be a memory detector to see selection.

So you could get to exoplanets and say that exoplanet

looks like there’s lots of selection going on there.

Maybe there’s evolution and maybe there’s going to be life.

So what I’m trying to say is narrow down

the regions of space where you say

there’s definitely evidence of memory as high assembly there

or not high assembly, because that would be life,

but where it’s capable of happening.

And then that would also help us frame the search for aliens.

I don’t know how likely it is to make the transition

to cells and all the other things.

I think you’re right.

But I think that we just need to get more data.

Well, I didn’t like the thought experiment

because I don’t like the idea

that if the universe has a maximal limit

on the amount it can generate per unit time

that our existence is actually precluding the existence

of other things.

Well, I’ll just say one thing.

But I think that’s probably true anyway

because of the resource limitations.

So I don’t like your thought experiment

because I think it’s wrong.

Well, no, no, I do like the thought experiment.

So what you’re trying to say is like,

there is a chain of events that goes back

that’s manifestly culminated with life on Earth.

And you’re not saying that life isn’t possible elsewhere.

You say that there has been these number of things,

contingent things that have happened

that have allowed life to merge here.

That doesn’t mean that life can’t emerge elsewhere,

but you’re saying that the intersection of events

may be concentrated here, right?

And I think there’s…

Not exactly.

It’s more like if you look at,

say the causal graphs are fundamental,

maybe space is an emergent property,

which is consistent with some proposals on quantum gravity,

but also how we talk about things in assembly theory.

Then the universe is causal graphs generating

more structure in causal graphs, right?

So this is how the universe is unfolding.

And maybe there’s a cap on the rate of generation.

Like there’s only so much stuff

that gets made per update of the universe.

And then if there’s a lot of stuff being made

in a particular region that happens

to look the same locally, spatially,

that’s an after effect of the fact

that the whole causal graph is updating.

Like it’s…

Yeah, I don’t know that.

I think that that doesn’t work.

I don’t think it works either,

but I don’t have a good argument in my mind about.

But I do like the idea of the capacity that universe,

cause you’ve got the number of states.

Yeah, we can come back to it.

Let me ask real quick.

Like why does different like local pockets

of the universe start remembering stuff?

How does memory emerge exactly?

So at the origin of the universe, it was very forgetful.

That’s when the physicists were happiest.

It was low memory objects,

which is like ultra low memory objects,

which is what the definition of stuff.

Okay, so how does memory emerge?

How does the temporal stickiness of objects emerge?

I’m gonna take a very chemocentric point of view

because I can’t imagine any other way of doing it.

You could think of other ways maybe.

But I would say heterogeneity in matter

is where the memory…

So you must have enough different ways

of rearranging matter for there to be a memory.

So what that means,

if you’ve got particles colliding in a box,

let’s just take some elements in a box.

Those elements can combine in a combinatorial set of ways.

So there’s a combinatorial explosion

of the number of molecules or minerals or solid objects,

bonds being made.

Because there’s such a large number,

the population of different objects that are possible,

this goes back to assembly theory

where assembly theory, there’s four types of universes.

So you’ve got basically, and this is what one was up earlier

where one universe where you’ve just got

everything is possible.

So you can take all the atoms

and combine them and make everything.

Then you’ve got basically what is the assembly combinatorial

where you basically have to accrue information in steps.

Then you’ve got assembly observed,

and then you’ve got the object assembly going back.

So what I’m trying to say is like,

if you can take atoms and make bonds,

let’s say you take a nitrogen atom and add it

to a carbon atom, you find an amino acid,

then you add another carbon atom on

in a particular configuration,

then another one, all different molecules.

They all represent different histories.

So I would say for me right now,

the most simple route into life seems to be

through recording memories and chemistry.

But that doesn’t mean there can’t be other ways

and can’t be other emergent effects.

But I think if you can make bonds

and lots of different bonds,

and those molecules can have a causal effect on the future.

So imagine a box of atoms,

and then you combine those atoms in some way.

So you make molecule A from load of atoms,

and then molecule A can go back to the box

and influence the box.

Then you make A prime or AB or ABC.

And that process keeps going,

and that’s where the memories come from,

is that heterogeneity in the universe from bonding.

I don’t know if that makes any sense.

And it’s beginning to flourish at the chemistry level.


So the physicists have no, like not enough.

Yeah, I mean.

They’re like desperately begging

for more freedom and heterogeneous components to play with.

Yeah, that’s exactly it.

What do you think about that, Sarah?

I mentioned already, I think it’s significant

that whatever physics governs life

emerges actually in chemistry.

It’s not relevant at the subatomic scale

or even at the atomic scale.

It’s in, well, atomic scale because chemistry.

But like when you get into this combinatorial diversity

that you get from combining things on the periodic table,

that’s when selection actually matters

or the fact that some things can exist

and others can’t exist actually starts to matter.

So I think of it like you don’t study gravity

inside the atomic nucleus.

You study it in terms of large scale structure

of the universe or black holes or things like that.

And whatever we’re talking about as physics of information

or physics of assembly becomes relevant

at a certain scale of reality.

And the transition that you’re talking about,

I would think of as just when you get a sufficient density

in terms of the assembly space

of like the relationship of the overlap

and the assembly space,

which is like a feature of common memory,

there is this transition to assembly dominated physics,

whatever that is.

Oh, like when we’re talking about,

and we’re trying to map out exactly

what that transition looks like.

We’re pretty sure of some of its features,

but we haven’t done all of the…

Do you think if you were there in the early universe,

you would have been able to predict

the emergence of chemistry and biology?

And I ask that because at this stage as humans,

do you think we can possibly predict the length of memory

that might be able to be formed later on

in this pocket of the universe?

Like how complex is, what is the ceiling of assembly?

I think as much time as you have in the past

is how much you can predict in the future.

Because it’s actually physical in the system

and you have to have enough time

for features of that structure to exist.

Wait, let me push back on that.

Isn’t there somewhere in the universe

that’s like a shortest path that’s been,

that stretches all the way to the beginning?


That’s building some giant monster?

Maybe, yeah.


So you can’t predict.

The universe has as much memory

as the largest assembly object in the universe.

Yeah. Right.

But so you can’t predict.

You can’t predict any deeper than that, no.


So like that, I guess what I’m saying is,

like what intuition do you have about complexity

living in the world that you’d have today?

Right, because you just, you can,

I mean I guess how long does it get more fun?

Like isn’t there gonna be at some point,

because there’s a heat death in the universe,

isn’t there going to be a point of the most,

of the highest assembly of object,

with the highest probability being generated?

When is the universe gonna be the most fun,

and can we freeze ourselves and then live then?


And will you know when you’re having the most fun

that this is the best time, you’re in your prime?

Are you going to do what everyone does,

which is deny that you’re in your prime,

and the best years are still ahead of you?

I don’t know.

What option do you have?

I don’t, I mean the problem is there’s lots of,

lots of really interesting features here.

I just wanna mention one thing that might be,

is I do think assembly theory applies all the way back

to subatomic particles.

And I also think that cosmological selection

might’ve been actually, there might’ve been,

I would say it’s a really boring bit,

but it’s really important for a cosmologist

that universes have gone through.

Was it Lee Smolin who proposed this?

Maybe that there is this,

that basically a universe evolves,

you’ve got the wrong constants, we’ll start again.

And the most productive constants

where you can allow particles to form in a certain way,

propagate to the next universe, and we go again.

So actually selection goes all the way back,

and there’s this cycle of universes.

And now this universe has been selected

because life can occur, and it carries on.

But I’ve really butchered that.

There is a much more.

So this is some aspect where through the selection process

there’s parameters that are being fine tuned,

and we happen to be living in one

where there’s some level of fine tuning.

Is there, given that, can you steel man the case

that we humans are alone in the universe?

We are the highest assembly index object in the universe.

Yeah, I can, I guess.

Sad though.

I mean, so from a.

Is it possible?

Yes, it’s possible.

Let’s assume.

Well, we know.

I mean, it’s possible.

So let me, so okay, so there is a particular

set of elements on Earth in a particular ratio,

and the right gravitational constant,

and the right viscosity, you know,

of stuff being able to move around,

the right distance from our sun,

right number of events where we have a moon,

the Earth is rotating.

The late heavy bombardment produced a lot of,

brought in the right stuff.

And Mars was cooking up, you know,

the right molecules first.

So it was habitable before Earth.

It was actually doing the combinatorial search.

And before Mars kind of became uninhabitable,

it seeded Earth with the right molecular replicators.

And there was just the right stuff on Earth,

and that’s how the miracle of life occurred.

Although I find I’m very uncomfortable with that

because actually, because life came so quickly

in the Earth’s past.

But that doesn’t mean that life is easy elsewhere.

It just might mean that,

because chemistry is actually not a long term thing.

Chemistry can happen quickly.

So maybe going on with the steel manning of the argument

to say actually, the fact that life emerged quickly

doesn’t mean that life is easy.

It just means that the chemistry was right on Earth,

and Earth is very special.

And that’s why there’s no life

anywhere else in the universe.

Yeah, so Sarah mentioned this kind of cascading thing.

So what if that’s the reason we’re lucky,

is that we got to have a rare cascading of,

like an accelerating cascading effect

in terms of the complexity of things.

So like, maybe most of the universe

is trying to get sticky with the memory,

and it’s not able to really form it.

And then we got really lucky in that.

And it has nothing,

like there’s a lot of Earth like conditions, let’s say,

but it’s just you really, really have to get lucky on this.

But I’m doing experiments right now.

In fact, experiments that Sarah and I are working on,

because we have some joint funding for this,

where we’re seeing that the universe

can get sticky really quickly.

Now, of course, we’re being very anthropocentric,

we’re using laboratory tools, we’re using theory,

but actually, the phenomena of selection,

the process of developing heterogeneity,

we can do in the lab.

We’re just seeing the very first hints of it.

And wouldn’t it be great if we can start to pin down

a bit more precisely becoming good Bayesianists for this,

for the origin of life and the emergence of life,

to finding out what kind of chemistries

we really need to look for.

And I’m becoming increasingly confident

we’ll be able to do that in the next few years.

Make life in the lab or make some selection in the lab

from inorganic stuff, from sand, from rocks,

from dead stuff, from moon.

Wouldn’t it be great to get stuff from the moon,

put it in our origin of life experiment,

and make moon life?

And restrict ourselves to interesting self replicating

stuff that we find on the moon.

Sarah, what do you think about this approach

of engineering life in order to understand life?

So building life in the machine.

Yeah, so, I mean, Lee and I are trying right now

to build a vision for a large institute

or experimental program, basically, to do this problem.

But I think of it as like, we need to simulate a planet.

So like the Large Hadron Collider was supposed

to be simulating conditions just after the Big Bang.

Lee’s built a lot of technology in his lab

to do these kind of selection engines.

But the question you’re asking is,

how many experiments do you need to run?

What volume of chemical space do you need to explore

before you actually see an event?

And I like to make an analogy

to one of my favorite particle physics experiments,

which is Super Kamiakande that’s looking

for the decay of the proton.

So this is something that we predicted theoretically,

but we’ve never observed in our universe.

And basically what they’re doing is every time

they don’t see a proton decay event,

they have a longer bound on the lifetime of a proton.

So imagine we built an experiment with the idea in mind

of trying to simulate planetary conditions,

physically simulate.

You can’t simulate origin life in a computer.

You have to do it in an experiment.

Simulate enough planetary conditions

to explore the space of what’s possible

and bound the probability for an origin life event.

Even if you’re not observing it,

you can talk about the probability.

But we, hopefully, life is not exponentially rare

and we would then be able to evolve

in an automated system alien life in the lab.

And if we can do that, then we understand the physics

as well as we understand what we can do

in particle accelerators.

So keep expanding physically the simulation,

the physical simulation, until something happens.

Yeah, or just build a big enough volume

of chemical experiments and evolve them.

So if you say volume, you mean like literally volume?

I mean physical volume in terms of space,

but I actually mean volume in terms

of the combinatorial space of chemistry.

So like.

How do you nicely control the combinatorial exploration,

the search space, such that it’s always like

you keep grabbing the low hanging fruit?

Yeah, how do you build a search engine for chemistry?

It’s like for aliens. I think you explained it really well.

We should carry on doing this.

I should pretend the physics, be the physicist,

you be the chemist.

So the way to do it is I will always play a joke.

Cause I like writing grants to ask for money

to do cool stuff.

But years ago I started wanting to build.

So I actually wanted the weather.

So I built this robot in my lab called the computer,

which is this robot you can program to do chemistry.

Now it’s a pro.

I made a programming language for the computer

and made it operate chemical equipment.

Originally I wrote grants to say,

Hey, I want to make an origin of life system.

And no one would give me any money for this.

They said, what this is ridiculous.

Why are you wanting to make, oh, it’s really hard.

It takes forever.

You’re not a very good origin of life chemist anyway.

Why would we give you any money?

And so I turned it around and said, can you,

can instead, can you give me money to make robots,

to make molecules are interesting.

And everyone went, yeah, okay, you can do that.

And that’s, so actually the funny thing is the computer

project, which I have in my lab, which is very briefly,

it’s just basically, it’s like literally an automated

test tube.

And we’ve made a programming language for the test tube,

which is cool, has come as literally came from this.

I went to my lab one day.

So I want to make a search engine to get origin of life

because they don’t have a planet.

And I thought about doing in a microfluidic format.

So microfluidic is very nano, very small channels

in device where you can basically have all the pipes

lit dump produced by lithography.

And you can have a chamber, maybe say between say 10

and a hundred microns in volume.

And we slot them all together like Lego,

and we can make an origin of life system.

And I could never get it to work.

And I realized I had to make, do chemistry at the kind of

test tube level and what you want to be able to do.

Yeah, it goes back to that tweet in 1981.

1981, the computer, we’re looking at a tweet from Lee.

In 1981, the computer was a distant dream in,

oh wow, this is the scientist looking back.

It is the young boy who dreamed.

In 2018, it was realized, spelled in a British way,

realized, which is the wrong way.

Yeah, I’m starting with Z, but not.

So now there’s a system that does the physical

manifestation or whatever the programming language,

the spec tells you to do.

Yeah, well in 1981, I got my first computer, ZX81.

What was the computer?


Sinclair ZX81.

It was, and I got a chemistry set.

And I liked the chemistry set and I liked the computer

and I just wanted to put them together.

I thought, wouldn’t it be cool if I could just use

the computer to control the chemistry set.

And obviously that was insane.

And I was like, you know, eight years old, right?

Nine years old, going on nine years old.

And then I invented the computer

just because I wanted to build this origin of life grid.

Which is like literally a billion test tubes

connected together in real time and real space,

basically throwing a chemical dice.

Throw dice, throw dice, throw dice.

You’re gonna get lucky.

And that’s what we, I think Sarah and I

have been thinking very deeply about.

Because, you know, there’s more money being spent

on the origin of the gravity

or looking at the Higgs boson than the origin of life, right?

And the origin of life is the, I think the biggest question

or not the biggest question, it is a big question.

Let’s put it that way.

It is the biggest question.

You’re okay saying that.

Okay, all right.

Isn’t it possible once you figure out the origin of life

that that’s not going to solve,

that’s not actually gonna solve

the question of what is life?

Because you’re kind of putting a lot of.

Yeah, I think they’re the same problem.

But you’re putting, is it possible

that you’re putting too many,

too much bets into this origin part?

Maybe the origin thing isn’t,

isn’t there always a turtle underneath the turtle?

Isn’t there a stack of turtles?

Because then if you create it in the lab,

maybe you need some other stuff.

Well, let’s not think about the origin.

Like in the lab, there’s still memory.

Yeah, yes.


So the experiment is already the product of evolution.

Right, in some maybe really deep way,

not an obvious way, in some very deep way.

So maybe the haters are always going to be like,

well, you have to reconstruct the fold.

You have to build a new script.

Fortunately for us, the haters are not aware

of that argument.

Well, no, I know, I just.

We’re the one making that argument usually, but yeah.

I just think that if we create life in the lab,

it’s not obvious that you’ll get

to the deep, deep understanding of necessarily,

what is the line between life and non life?

No, I think, so there’s so much here.

I’m just like playing devil.

So much here, but let me play devil’s advocate

back in a previous conversation, right?

And say, yeah, I will.

Why not?

We’ve got time.

Yeah, let’s go.

Cellular automata.

Cellular automata, these very, very simple things

where you color squares black or white

and implement rules and play them in time.

And you can get these very, very complex patterns coming out.

You know, there’s nice rules.

There are Turing complete rules and I would argue

that cellular automata don’t really exist on their own.

They have to exist in a computing device.

If that, well, that’s computing devices,

a piece of paper and abstraction,

a mathematician drawing a grid or a framework.

Now, so I would argue CAs are beautiful things,

simple, going complex, but the complexity is all borrowed

from the lithography, the numbers.

Right, now let’s take that same argument

with the chemistry experiment origin of life.

What you need to be able to do is go,

and I’m inspired to do this,

to go out and look for CAs occur in nature.

You know, let’s kind of, let’s find some CAs

that just emerge in our universe and.

For people just, sorry to interrupt,

for people just listening and in general,

I think what we’re looking at is a cellular automata

where again, as Lee described,

there is just binary black or white squares

and they only have local information

and they’re born and they die.

And you would think nothing interesting would emerge,

but actually what we’re looking at is something

that I believe is called glider guns

or a glider gun, which is moving objects

in this multi cell space that look like they’re organisms

that have much more information,

that have much more complexity

than the individual building components.

In fact, look like they have a long term memory

while the individual components don’t seem like

they have any memory at all, which is fascinating.

The argument here is that has to exist

on all this layer of infrastructure, right?

And though it looks simple.

And then what I would make, the argument I would make

if I were you, say, well, I think CAs are really simple

and everywhere, is say, show me how they emerge

in a substrate.

Now this goes to the origin of life, machine.

I don’t think we want to do the origin of life,

just any origin is good.

So we do, so we literally have our sand shaker,

shake the sand like massive grid of chemistry experiments,

shaking sand, shaking whatever.

And then because we know what we’ve put in,

so we know how we’ve cheated and the same way with CA,

we know how we’ve cheated,

we know the number of operations needed,

we know how big a grid we want to get this.

If we could then say, okay,

how can we generate this recipe in the lab

and make a life form?

What contingency did we need to put in?

And we’re upfront about how we cheated, okay?

Say, oh, you had to shake it, it was a periodic,

planet rotates, it’s tried, comes in and out.

So, and then we can start to basically say,

okay, how difficult is it for these features to be found?

And then we can look for extra planets and other features.

So I think Sarah is absolutely right.

We want to explain to people we’re cheating.

In fact, we have to cheat.

No one has given, I’m good at writing grants,

well, I used to be, I’m not very good right now,

I keep getting rejected,

but writing a grant for a planet in 100 million years,

no grant fund there is going to give me that,

but maybe money to make a kind of a grid,

a computer grid, origin of life computer grid.

In physical space.

In physical space, and just do it.

So Sarah said something which is,

you can’t simulate the origin of life in a computer,

so like in simulation, why not?

What are your, you said it very confidently,

so is it possible and why would it be very difficult?

Like what’s your intuition there?

I think it’s very difficult right now

because we don’t know the physics,

but if you go based on principles of assembly theory

and you think every molecule is actually

a very large causal graph, not just the molecule,

then you have to simulate all the features

of those causal graphs,

and I think it becomes computationally intractable.

You might as well just build the experiment.

Because you have, in the physical space,

you have all the objects with all the memories.


In the computer, you would have to copy them

or reconstruct them.


Yeah, that’s a beautifully put,

and I would say that lots of people,

you just don’t have enough resources.

It’s easier to actually do the physical experiment

because we are literally,

I would view the physical experiment

almost like a computational experiment.

We’re just outsourcing, it’s just basically,

we’re just outsourcing all the matrix algebra.

And on your point about the experiment

being also an example of life,

it’s almost like you want to design,

it’s like all of us are lineages

of propagating information across time,

and so everything we do becomes part of life

because it’s part of that causal chain.

So it’s like you want to try to pinch off

as much as you can of the information

from your causal chain that goes into the experiment,

but you can’t pinch off all of it

to move it to a different timeline.

It’s always going to be part of your timeline.

But at least if you can control

how much information you put in,

you can try to see how much does that particular trajectory

you’ve set up start generating its own assembly.

So you know where it starts,

and then you want to try to see it take off on its own

when you try to pinch it off as much as possible.

Got it.

Quick pause, bathroom break.


All right, cool.

And now we’re back.

All right.

We talked about the early days of the universe

when there was just stuff and no memory,

not even causality.

I think Lee at least implied

that causality is immersion somehow.

We could discuss this.

What happened before this all originated?

What’s outside the universe?

Divided by zero.

Okay, so it’s not relevant, not understandable.

Is it useful to even ask the question?


Just because it’s so hard?

No, it’s not hard.

It’s just not a question.

If I can’t do an experiment or even think of an experiment,

the question doesn’t exist.

Well, no, you can’t think of a lot of experiments,

no offense.

What I mean is I can’t.

Your causality graph is like,

this is what we’re talking about.

It’s like there is limits to your ability

to construct experiments.

I agree, but I was trying to be facetious

and I’m trying to make a point

because I think that if there is a causal bottleneck

through which information can’t propagate in principle,

then it’s very hard to think of an experiment,

even in principle, even one that’s beyond

my mediocre intellect, right?

Which is fine.

I’m happy to accept that.

But this is one of the things I actually do think

there was something before the Big Bang

because I would say that I think the Big Bang

just couldn’t occur and create time.

Time created the Big Bang.

So there was time before the Big Bang.


There was no space, but there was time.


But I mean, I’m just making that stuff up

just to make all the physicists happy,

but I think it’s…

Do you think that would make them happy

because they would be quite upset, actually.

And why would they be upset?

Because they would say that time can’t exist

before the Big Bang.

Yeah, I mean, this goes back to an argument

that you might not want to have the argument here.

I was talking to Sarah earlier today

about an argument we had about time a long time ago.


A long time in time.

And what I would, it’s like, I think there is this thing

called time or state creation.

The universe is creating states and it’s outside of space,

but they create space.

So what I mean is you can imagine there are states

being created all the time.

And there is this thing called time.

Time is a clock, which you can use to measure

when things happen, but that doesn’t mean,

because you can’t measure something,

that states aren’t being created.

And so you might locally refer to the Big Bang

and the Big Bang occurred at some point

when those states were there.

Probably there had to be enough states

for the Big Bang to occur.

And then, but I think that there is something wrong

with our conception of how the universe was created

and the Big Bang because we don’t really get time.

Because again, I don’t want to become boring

and sound like a broken record, but time is a real thing.

And until I can really explain that more elegantly,

I’m just gonna get into more trouble.

Well, we’re gonna talk about time

because time is a useful measuring device for experiments,

but also time is an ideal, okay.

But let me first ask Sarah, what do you think?

Is it a useful question to ask what happened

before the Big Bang?

Is it a useful question to ask what’s outside the universe?

So I would think about it as the Big Bang

is an event that we reconstructed

as probably happening in the past of our universe

based on current observational data.

And so the way I like to think about it

is we exist locally in something called the universe.

So, and going back to like the physics of existence,

we exist locally in the space of all things that could exist

and we can infer certain properties of the structure

of where we exist locally.

And one of the properties that we’ve inferred in the past

is that there is a thing we call the Big Bang.

There’s some signatures of our local environment

that indicate that there was a very low information event

that started our universe.

I think that’s actually just an artifact

of the structure of the assembly space

that when you start losing all the memory in the objects,

it looks like what we call a Big Bang.

So I think it makes sense to talk about

where you are locally.

I think it makes sense to talk about

counterfactual possibilities,

what could exist outside the universe

in the sense that they become part of our reasoning

and therefore part of our causal chain

of things that we can do.

So like the multiverse in my mind exists,

but it doesn’t exist as a multiverse

of possible universes.

It exists as an idea in our minds

that allows us to reason about how physics works

and then to do physics differently

because we reason about it that way.

So I always like to recenter it on things exist,

but they don’t always exist like we think they exist.

So when we’re thinking about things outside the universe,

they absolutely exist because we’re thinking about them,

but they don’t look like the projections in our minds.

They’re something else.

And something you said just gave an idea

to go back to your question.

If there was, I mean, if something caused the Big Bang,

if there was some memory or some artifact of that,

then of course, to answer your question,

it’s worth going back to that

because that would imply there is something

beyond that barrier, that filter.

And that’s what you were saying, I guess, right?

I’m agnostic to what exists outside the universe.

I just don’t think that.

I think the most interesting things for us to be doing

are finding explanations that allow us to do more,

like that optimism.

So I tend to draw the boundary on questions I ask

as being scientific ones because I find

that that’s where the most creative potential is

to impact the future trajectory

of what we’re doing on this planet.

It’s an interesting thing about the Big Bang

is basically from our current perspective

of what we’re able to detect,

it’s the time when things were forgotten.


It’s the time to reset from our limited perspective.

And so the question is, is it useful to ever study

the thing that was forgotten?

Or should we focus just on the memories

that are still there?

Well, the point I was trying to make about the experiment

is I was trying to say both things.

And I think perhaps yes, from the portfolio point of view,

if you could then imagine what was forgotten

and then work forwards,

you will have different consequences.

So then it becomes testable.

So as long as we can find tests,

then it’s definitely worth thinking about.

What I don’t like is when physicists say

what happened before the Big Bang

and before, before, before,

without giving me any credible conjecture

about how would we know the difference?

But the way you framed it is quite nice.

I like that.

It’s like, what have we forgotten?

Is there room for God in assembly theory?

Who’s God?

I like arguments for a necessary being better than God.

Well, I think I said it earlier.

What’s a necessary being?

What’s a necessary?

Like something that has to exist.

Oh, so you like, I mean, you like the shortest path.

Like does God need?

No, no, no.

I mean, well, you can go back to like Thomas Aquinas

and arguments for the existence of God.

But I think most of the interesting theological arguments

are always about whether something has to exist

or there was a first thing that had to exist.

But I think there’s a lot of logical loopholes

in those kind of arguments.

Well, so God here, meaning the machine

that creates, that generates the stuff.

But God, so what I was trying to say earlier is that.

Isn’t that just the universe though?

Yeah, yeah.

Well, yeah, well, but there’s a difference between,

I sort of imagine like a black box, like a machine.


I mean, I would be more comfortable calling that God

because it’s a machine.

You go into a room and there’s a thing with a button.

Yeah, I don’t like the great programmer in the sky version.

Yeah, but if it’s more kind of like,

I don’t like to think of, if you look at a cellular automata,

if it’s the cells and the rules,

that doesn’t feel like God

that generates a bunch of stuff.

But if there’s a machine like that does,

that runs the cellular automata and set the rules,

then that feels like God.

That sort of, in terms of terminology.

So I wonder if there’s like a machine

that’s required to generate this universe.

That’s very sort of important for running this in the lab.

So as I said earlier, I think I said this earlier,

that I can’t remember the phrase, but something like,

I mean, does God exist in our universe?


Where does God exist?

God at least exists in abstraction in our minds,

particularly of people who have religious faith

they believe in.

But let’s then take, but you’re talking a little bit more

about generics, say, well,

is there a mechanism beyond the universe you’re calling God?

I would say God did not exist at the beginning,

but he or she does now.

Because I’m saying the mechanism.

Well, you don’t know that he didn’t exist in the beginning.

So like this could be us in our minds trying to,

like just listening to gravitational waves,

detecting gravitational waves.

It’s the same thing.

It’s us trying to go back further and further

into our memories to try to understand the machines

that make up, that make up us.

And so it’s possible that we’re trying to grasp

at possible kind of what kind of machines could create.

There’s always a tweet.

If the universe is a computer, then God must have built it

because computers need creators.

There you go.

And then Joshe Bach replied,

since there’s something rather than nothing,

perhaps existence is the default.

If existence is the default, then many computers exist.

Creator gods are necessary computers,

unnecessarily computers too.

I’m very confused by that, but that’s an interesting idea

that existence is the default versus nonexistence.

I agree with that, but the rest is not.

And then Lee responds,

perhaps this reasoning is incomplete.

That’s how scientists talk trash each other

on Twitter apparently.

Which part don’t you agree with?

When he said if existence is default,

then many computers exist,

this comes back to the inventor and discovery argument.

I would say the universe at the beginning wasn’t capable

of computation because there wasn’t enough technology,

enough states.

So what you’re saying is if God is a mechanism,

so I might actually agree,

but then the thing is lots of people see God

as more than a mechanism.

For me, God could be the causal graph in assembly theory

that creates all the stuff that the memories we know.

And the fact that we can even relate to each other

is because we have the same, we share that heritage.

And why we love each other

or we like to see God in each other

is it’s just we know we have a shared existence.

So if the God is the mechanism

that created this whole thing,

I think a lot of people see God in a religious sense

as that mechanism also being able to communicate

with the objects it creates.

And if it’s just the mechanism,

we won’t be able to communicate with the objects it creates.

It can only create.

You can’t interact with the…

Well, there’s versions of God that create the universe

and then left.

Yeah, like spark.

For some religions.

The first spark, yeah.

But I think I liked your analogy

of the machine and the rules, right?

But I think part of the problem is,

I mean, we have this conception

that we can disentangle the rules

from the physical substrate, right?

And that’s the whole thing about software and hardware

being separate or the way Newton wrote his laws

that there was some,

like they exist outside the universe.

They’re not actually a feature of the universe.

They don’t have to emerge out of the universe itself.

So I think if you merge your two views,

then it gets back to the God is the universe.

And then I think the deeper question

is why does it seem like there’s meaning and purpose?

And if I think about the features of the universe

that give it the most meaning and purpose,

those are what we would call

the living components of the universe.

So if you wanted to say God is a physically real thing,

which you were saying

is like an emergent property of our minds,

but I would just say the way the universe

creates meaning and purpose,

there is really a physics there.

It’s not like a illusory thing.

And that is just what the physics of life is.

Is it possible that we’ve forgotten

much of the mechanisms that created the universe?

So like, so basically, you know,

whatever, if God is that mechanism,

we just leave parts of that behind.

Well, but the universe is constantly generating itself.

So if God is that mechanism,

it would be that that would still be active today.

I don’t, like, I’m agnostic,

but if I recall the things I believe in God

in the way that some people talk about God,

I would say that God is, you know,

like in the universe now, it’s not an absent thing.

So I think there’s a mislabeling here

because you’re, I mean, I mean, I’m a professional idiot,

actually, but, but, um.

You should put that on your CV.


Professionally, not recreationally or amateur,

but professionally, you’re paid for it.

I would say if you were talking about God,

I mean, again, I’m way out, way out of my depth here,

and I almost feel uncomfortable.

Yeah, but I feel quite uncomfortable articulating,

but I’ll try.

For me, a lot of people that think of God as a consciousness

and a reasoning entity that actually has causal power,

and you’re, and so you’re, it’s like,

then you’re saying like gravity could be God

or time could be God.

I mean, I think for me, for my conception of time

is probably as fundamental as God

because it gave rise to human intelligence and consciousness

in which we can have this abstract notion of God.

So I think that you’re maybe talking about God

in a very mechanistic, kind of unsophisticated sense,

whereas other people say that God is more sophisticated

and got all this, you know, feelings and love

and, you know, and this abstracting ability.

So is that what, or do you mean that?

Do you mean God as in this conscious entity

that decided to flick the universe into existence?

Well, one of the features that God would have

is the ability to flick the universe into existence.

I, you know, like Windows 95,

I don’t know if God is Windows 95 or Windows XP

or Windows 10, I don’t know the full feature set.

So at the very least, you have to flick the universe

into existence, and then other features might include

ability to interact with that universe in interesting ways,

and then how do you interact with the universe

in interesting ways?

You have to be able to speak the language

of its different components.

So in order to interact with humans,

you have to know how to act humanlike.

So I don’t know, but it seems like

whatever mechanism created the universe

might want to also generate local pockets of mechanisms

that can interact with that.

Like inject.

Like God was lonely?

Yeah, I mean, it could be just a teenager

and another just playing a video game.

Yeah, maybe.

Well, I was gonna say, I mean, I don’t,

so this is referring to our origin of life engine.

It’s like, I don’t believe in God,

but that doesn’t mean I don’t wanna be one.


I don’t wanna make a universe and make a life form,

but that may be rude to people who have

dear religious beliefs.

What I mean by that is if we are able to create

an entirely new life form, different chemistry,

different culture, what does it make us?

By that definition, it makes us gods, right?

Well, there is.

I mean, like when you have children,

you’re like one of the magical things of that

is you’re kind of mini gods.

I mean, first of all, from a child’s perspective,

parents are gods for quite a while.

And then, I mean, in the positive sense,

there’s a magic to that.

That’s why I love robotics,

is you instill life into something,

and that makes you feel godlike in a sort of positive way.

Being a creator is a positive feeling.

Creator, yeah, exactly, on a small scale.

And then God would be a creator

at the largest possible scale, I suppose.

Okay, you mentioned offline the Assembletron.



What’s an Assembletron?

This is an early idea of something you’re thinking about.

So Sarah’s team, well, I think Sarah’s team

are interested in using AI to understand life.

My team is, and I’m wondering if we could apply

the principles of assembly theory,

that is the causal structure that you get

with assembly theory, and hybridize it,

and make a new type of neuron, if you like.

I mean, there are causal neural networks out there,

but they are not quite the architecture

of what I would like.

I would like to associate memory bits with,

basically, I’d like to make a,

rather than having an ASIC for neural networks,

I wanna make an ASIC for assembly networks, right?


So can you say that again?

Assembly networks.

So what is a thing with an input and an output,

and it’s like a neural network type of thing,

what does it do exactly?

What’s the input, what’s the output?

So in this case, so if you’re talking about

a general neural network, I mean,

in general neural network, you can train it

on any sort of data, right, depending on the framework,

whether it’s like text, or image data, or whatnot.

And that’s fine, but there’s no causal structure

associated with that data.

Now just imagine, rather than, you know,

let’s say we’re gonna classify a difference

between cat and dog, right?

Classic cat and dog neural network.

What about if the system understood the assembly space,

it created the cat and the dog,

and rather than guessing what was happening,

and training on those images,

and not understanding those features,

you almost like, you could imagine doing a,

going back a step and doing, and training,

going back a step and doing the training,

going back a step, back a step, back a step,

and I wonder if that is actually the origin of intelligence,

or how we’ll crack intelligence.

Because we need to, because we’ll create

the entire graph of events, and be able to kind of

look at cause and effect across those graphs.

I’m explaining it really badly,

but it’s a gene of an idea, and I’m guessing

very smart, very rich people in AI are already doing this.

Trying to not generate cats and dogs,

but trying to generate things of high assembly index.

Yeah, and I think, and also using causal graphs

in neural networks, and machine learning, and deep learning,

maybe building a new architecture.

I’m just wondering, is there something

we can get out of assembly theory

that allows us to rebuild current machine learning

architectures to give causation more cheaply?

I mean, I don’t know if that’s what you,

we’ve been inventing this for a little while,

but we’re trying to finish the theory paper first

before we do anything else.

Yeah, you also wanna have, say, goal directed behavior

in neural networks, then assembly theory

is a good framework for doing that.

Daniel’s been thinking about that a lot.

And I think it’s a really interesting idea

that you can map concepts from how neural networks learn

to thinking about goal directed behavior

as a learning process.

That you’re learning a specific goal.

The universe is learning a goal

when it generates a particular structure,

and that you can map that physical structure

in a neural network.

What’s the goal?

Well, in a neural network, you’re designing the goal.

In biology, I mean, people are not supposed to use

teleological language in biology, which is ridiculous,

but, because goals are real things.

They’re just post selected.

So you can talk about goals after the fact.

Once a goal emerges in the universe,

that physical entity has a goal.

But Lee and I came up with a test for,

like a Turing test for goal directed behavior

based on the idea of assembly.

We have to formalize it still,

but I would like to write a paper on it.

But the basic idea is if you had two systems

that were completely equivalent,

you know, like in the instantaneous,

like physical experimental setup.

So Lee has to figure out how to do this.

But there was something that would be different

in their future.

And there was a symmetry breaking you observe

in the present based on that possibility

of that future outcome.

Then you could say that that system

had some representation of some kind of goal in mind

about what it wanted to do in the future.

And I, so goals are interesting

because they don’t exist as instantaneous things.

They exist across time,

which is one of the reasons that assembly theories

may be more naturally able to account

for the existence of goals.

So goals are, they only exist in time,

or they manifest themselves in time

through, you said, symmetry breaking.

So it’s almost like, imagine,

like if representations in your mind are real, right?

And you can imagine future possibilities,

but imagine everything else is physically equivalent.

And the only thing that you actually change

your decision based on is what you model

as being the future outcome.

Then somehow that representation in your mind

of the future outcome becomes causal

to what you’re doing now.

So there’s kind of like retro causal effect,

but it’s not actually retro causal.

It’s just that your assembly space

is actually includes those possibilities

as part of the structure.

It’s just, you’re not observing all the features

of the assembly space in the current moment.

Or the possibilities exist, but they don’t become a goal

until they’re realized.

So one of the features of assembly space

that’s super interesting,

and it’s easier to envision with like Legos, for example,

is if you’re thinking about an assembly space,

you can’t observe the entire assembly space

in any instant in time.

So if you imagine a stack of Legos,

and you wanna look at the assembly space of a stack of Legos,

you have to break the Legos apart,

and then you look at all the possible ways

of building up the original object.

So now you have in your mind

the goal of building that object,

and you have all the possible ways of doing it.

And those are actual physical features of that object,

but that object doesn’t always exist.

What exists is the possibility of generating it.

And the possibilities are always infinite.

Well, for that particular object,

like it has a well defined assembly space.

And I guess what I’m saying is that object

is the assembly space,

but you actually have to unpack that object across time

to view that feature of it.

It’s only an observable across time.

The term goal is such a important

and difficult to explain concept, right?

Because what you want is a way is like,

I think only conscious beings can have conscious goals.

Everything else is doing selection.

And but selection does invent goals.

And in a way that the way that biology reinterprets the past

in the present is kind of helped you to understand

there was a goal in the past now, right?

It’s kind of like goals only exist back in time.

So first of all, only conscious beings

can have conscious goals.

I’m not even gonna touch that one.


Go for it, come on.

The line between conscious goals

and non conscious goals, exactly.

And also maybe just on top of that,

you said a Turing test for goal directed behavior.

What does a Turing test potentially look like?

So if you’ve got two objects, we were thinking about this.

So we actually got some funding

to work to go on two teams.

So I’m trying to do, and part of this

is I’m trying to do a bit of theory

and Sarah is teaching me a bit of theory

and Sarah is trying to design experiments

and I’m teaching experiments.

Cause I think it’s really good for us to have that.

So say, when would a, so that’s good.

I like this, I’m sure we’re using Dan Dennett essay.

Yeah, and I can explain why we wouldn’t want to call it

a Turing test after, but Dan Dennett.

So Dan Dennett wrote this really nice essay

about herding cats and freewill inflation.

The title is so brilliant.

That’s the actual title?

That’s the title, yeah.

Herding cats and freewill inflation.

Yeah, something like that.

I mean, it’s not, maybe not.

And so.

No, I think that’s right.

So if you’ve got a, let’s imagine you’ve got two objects

on a hillside, okay?

And it just happens to be a snowy hill

and let’s just say you see an object

get rolling down the hill.

Or you, and the rock rolls down the hill

but the start goes to the end.

How do you know that object’s had a goal?

Now you unveil the object

and you’ll see it’s actually a skier.

And the skier starts at the top and goes down the bottom.


Then you look at the rock.

Rock rolls down the hill and goes to the bottom.

How can you tell the difference between the two?

So, and what Dan says is like, well, this is clear.

The skier’s in control.

And because they’re adjusting the trajectory,

so there’s some updating going on.

Then the only way you could really do that

is if you put the skier back to the top of the hill again,

they would tend to start roughly in the same space

and probably take all that complex set of trajectories

and end up pretty much at the same finish point, right?

With plus or minus a few meters.

Whereas if it was just a random rock

going down to a random trajectory, that wouldn’t happen.

And so what Sarah and I were kind of doing

when we were writing this grant,

we were like, we need to somehow instantiate the skier

and the rock in an experiment.

And then say, okay, when does the object,

so for an object to have a goal, it has to have an update.

It has to have some sensing and some kind of,

you know, inbuilt actuation to respond to the environment.

And then we just have to iterate on that.

And maybe Sarah, you can then fill in the Turing test part.

Well, yeah, I guess the motivation for me

was slightly different.

So I get really frustrated about conversations

about consciousness as most people do.

You know, a lot of people are,

which is not necessarily related to free will directly

or to this goal directed behavior.

But I think there’s a whole set of bundled

and related topics here.

But I think for me, I was, you know,

everybody’s always interested in explaining

intrinsic experience and quantifying intrinsic experience.

And there’s all sorts of problems with that

because you can never actually be another physical system.

So you can’t know what it’s like

to be another physical system.

So I always thought there must be some way

of getting at this problem about

if an agent or an entity is conscious

or at least has internal representations

and those are real physical things,

that it must have causal consequences.

So the way I would ask the question of consciousness

is not, you know, what is it like intrinsically?

But if things have intrinsic experience,

is there any observable difference from the outside

about the kind of causation

that that physical system would enact in?

And for me, the most interesting thing that humans do

is have imagination.

So like we can imagine rockets centuries

before we build them.

They’ve become real physical things

because we imagine them.

And people might disentangle that from conscious experience,

but I think a lot of the sort of imagination we do

is actually a conscious process.

So then this becomes a question of

if I were observing systems and I said,

one had an internal representation,

which is slightly different

than a conscious experience, obviously.

So I’m entangling some concepts,

but it’s a loose set of thought experiments.

Then how, and I set them up

in a physically equivalent situation.

Would it be the case that

there would be experimental observables associated with it?

And that became the idea of trying to actually

measure for internal representation and conscious.

So Turing basically didn’t wanna do that.

You just wanted the machine that could emulate

and trick you into having the behavior,

but never dealt with the internal experience

because he didn’t know how to do that.

And I guess I was wondering,

is there a way to set up the experiment

where you could actually test for that?

For imagination that led to the thing.

That there was something internal going on,

some kind of inner world, as people say,

or you could say, it actually is an agent,

it’s making decisions, it has an internal representation.

And whether you say that’s experience or not

is a different thing, but at least the feature

that there’s some abstraction it’s doing

that’s not obvious from looking at the physical substrates.

Do you think it’s possible to do that kind of thing?

One of the compelling things about the Turing test

is that defining intelligence,

defining any complicated concept

as a thing like observing it from the surface

and not caring about what’s going on deep inside

because how do you know?

That’s the point.

So the idea is exactly that.

So what we’re trying to do,

the Turing test for goal directedness

is literally take some objects

that clearly don’t have any internal representation,

grains of sand blowing on the beach or something,

and I don’t know, a crab wandering around on the beach

and then generating an experiment

where literally the experiment generates an entity

that literally has no internal representation to sand,

these are oil droplets actually,

what we’ve got in mind, a robot that makes oil droplets.

But then what we wanna try and do

is train the oil droplets to be like crabs,

give them an internal representation,

give them the ability to integrate information

from the environment so they remember the past,

are in the present, and can imagine a future.

And in a very limited way, their kind of game engine,

their limited simulation of the world

allows them to then make a decision.

They’re objects across time.

So then you would run a bunch of crabs

like over and over and over and over?

How many crabs, Lee?

How many, is there, what’s,

because you have to have a large number of crabs,

what does your theory say, is there a mathematical?

We’re working on it, I mean, this is literally.

Limit, crab limit.

There’s literally a.


There’s literally a.

What’s the herding cats have to do?

Oh, that’s random, wait, what’s cats,

in the title by Daniel Dennett,

Herding Cats and the Free Will Inflation.

So, I love this.

What does herding cats mean,

what does free will inflation mean?

So this, I love this essay,

because it explained to me

how I can live in a deterministic universe,

but have, not free will, but have freedom.

And also it helped me explain

that time needed to be a real thing in this universe.

So what basically Dan was saying here is like,

how do you, how do these cats appear

to just do what they want, right?

And if you live in a deterministic universe,

why do the cats do these things?

You know, aren’t they just, isn’t it all obvious?

And how does free will inflate the universe?

And for me, I mean, probably I love the essay

because my interpretation of the essay

in assembly theory makes complete sense.

Because you need an expanding universe

in assembly theory to create novelty

that you search for,

that then when you find something interesting

and you keep doing it because it’s cool

and it gives you an advantage,

then it appears in the past to be a goal.

So what does, in assembly theory,

the expansion of the universe look like?

What are we talking about?

Why does the expansion of the universe

give you more possibilities of novelty and cool stuff?

So for me, I don’t think about the universe

in terms of big bang and space.

I think about it in terms of the big memory expansion.

That you have one, you only have the ability

to store one bit of information,

so then you can’t do very much.

So what the universe has been doing since forever,

it’s been creating more,

it’s been increasing the size of its RAM, okay?

So it’s like one megabyte, two megabyte,

three megabyte, four megabytes, all the way up.

And so the more RAM you have,

the more you can remember about the past,

which allows you to do cooler things in the future.

So if you can remember how to launch a rocket,

then you might be able to imagine how to land a rocket,

and then relaunch, reland, and carry on.

And so you’re able to expand the space

and remember the past.

And so that’s why I think it’s very important.

But not a perfect memory.

It’s an interesting question,

whether there’s some forgetting that happens

that might increase.

Is the expansion of the forgetting, at some point,

accelerate faster than the remembering?

I think that that’s a very important thing

that probably intelligence does,

and we’re gonna learn in machine learning about,

because you want machine learning right now,

or artificial intelligence right now,

doesn’t have memory right,

but you want the ability to,

or not for, if you want to get to human like consciousness,

you need to have the ability, I suppose, to remember stuff

and then to selectively forget stuff

so you can re remember it and compress it.

Arguably, the way that we come up with new physical laws.

I think that there is a great deal to be gained

from having the ability to remember things,

but then when you forget them,

you can then have a,

you can basically do the simulation again

and work out if you get to that compressed representation.

So that’s in cycles.

So cycles of remembering and forgetting

are probably important,

but there shouldn’t be excuse to have a universe

with no memory in it.

The universe is gonna remember that it forgot,

but just not tell you.

I’m looking at this paper

and it’s talking about a puppet controlling a puppet

controlling a puppet controlling a puppet controlling

a puppet controlling a puppet,

conceptually easy to understand,

but physically impossible,

as physically impossible as predicting a fair coin toss.

I don’t know what he’s talking about,

but there’s pictures of puppets controlling puppets.

Let me ask you,

there’s a few things I wanna ask,

but we brought up time quite a bit.

You guys tweet about time quite a bit.

What is time in all of this?

We kind of mentioned it a bunch.

Is it not important at all in terms of,

is it just a word?

Should we be talking about causality mostly?

Like Sarah, what do you think?

Is, we’ve talked about like memories.

Is that the fundamental thing

that we should be thinking about?

And time is just a useful measurement device or something

like that.

Well, there’s different concepts of time, right?

So I think in assembly theory,

when we’re talking about time,

we’re talking about the ordering of things.

So that’s the causal graph part.

And so then the fundamental structure of the universe

is that there is a certain ordering

and certain things can’t happen till other things happen.

But usually when we colloquially talk about time,

we’re talking about the flow of time.

And I guess Lee and I were actually debating

about this this morning.

So in talking on it, walking on the river here,

which is a very lovely spot for talking about time,

but that when the universe is updating,

it’s transitioning between things that exist now

and things that exist now.

That’s really the flow of time.

So you have to separate out those concepts at bare minimum.

And then there’s also an arrow of time

that people talk about in physics,

which is that time doesn’t appear to have a directionality

in fundamental physics, but it does to us, right?

Like we can’t go backwards in time.

And usually that would be explained in physics

in terms of, well, there’s a cosmological arrow of time,

but there’s also the thermodynamic arrow of time

of increasing entropy.

But what we would say in assembly theory

is that there is a clear directionality.

The universe only runs in one direction,

which is why some things, it’s easy to make,

if the universe runs in one direction,

it’s easy to make processes look reversible.

For example, if they have no memory,

they’re easy to run forward and backwards,

which is why the laws of physics that we have now

look the way they do, because they involve objects

that have no memory.

But when you get to things like us,

it becomes very clear that the universe

has a directionality associated to it.

So it’s not reversible at all.

It’s the no man ever steps in the same river.

I just have to bring that out

because you walked on the river.

No man ever steps in the same river twice

before he’s not the same river and he’s not the same man.

So it’s not reversible, any of the same.

No, but reversibility is an emergent property, right?

So we think of the reversibility laws as being fundamental

and the irreversibility as being emergent.

But I think what we would say from how we think about it,

and it certainly seems to be the case

for our perception of time,

but also what’s happening in biological evolution,

you can make things reversible,

but it requires work to do it.

And it requires certain machines

to run it forward and backward.

And Chiara Marletto is working some interesting ideas

on constructor theory related to that,

which is a totally different set of ideas.

You can travel back in time sometimes.


You can’t travel actually back in time,

but you could reconstruct things

that have existed in the past.

You’re always moving forward in time,

but you can cycle through.

Like, I mean, I can…

Can I clarify what you just said?

Yeah, yeah, go for it.

Quickly, you travel forward in time to travel back.


Thank you, that really clarified it.

What Sarah’s saying is you don’t go back in time,

you recreate what happened in the past in the future

and inspect it again.

So in that local pocket of time,

it’s as if you travel back in time.

So I don’t…

How is that not traveling back in time?

Because you’re not going back

to your same self back in time.

You’re creating that in the future.

But everything else is the same as it was in the past.

No, no, no, no, it’s not in registry.

I mean, it goes back to the big question I’m saying.

I mean, this is something I was trying to look up today

when we first had this discussion

and I was talking to Sarah on Skype and said,

by the way, time is the fundamental thing in the universe.

She almost hung up on me.

Right, but you can even…

I mean, if you wanna make an analogy to computation,

and I think Charles Bennett actually has a paper on this

like about reversible computation

and reversible Turing machines.

In order to make it reversible,

you have to store memory to run the process backwards.

So time is always running forward in that

because you have to write the memory.

You can’t erase the memory.

You can erase the memory,

but the point when you go back to zero, right?

But the whole point is that in order to have a process

that even runs in both directions,

you have to start talking about memory

to store the information to run it backwards.

I got it.

So you can’t really then…

You can’t have it exactly how it was in the past.

Yeah, exactly.

You have extra stuff, extra baggage always.


A really important thing that I want to say on this,

I think if I try and get it right,

I have to say that if you can think

that the universe is expanding

in terms of the number of boxes

that it has to store states, right?

And this is where the directionality of the universe

comes from, everything comes from.

You could erase what’s in those boxes,

but the fact you’ve now got so many boxes at time now

in this present, there’s more of those boxes

than there were in the past.

See, but the boxes aren’t physical boxes.

They’re not space or time.

Why is the number of boxes always expanding?

It’s very hard to imagine this

because we live in space.

So what I’m saying, which is I think probably correct,

is that we just, let’s just imagine for a second,

there is a nonlocal situation,

but there are these things called states

and that the universe irrespective

of whether you measure anything,

there is a universal, let’s call it a clock

or a state creator.

Maybe we can call it, that’s why maybe you can call it God,

but let’s call it a state creator

where the universe is expanding

in the number of states it has.

Why are you saying it’s expanding though?

Is that obvious that it’s expanding?

It’s obvious because that’s where the,

because we, we.

That’s a source of novelty.

It’s a source of novelty

and it also explains why the universe is not predictable.

How do you know it’s not predictable?

I just like interrupting you.

Sorry, it’s fun, because you’re struggling.

I’m struggling because I’m trying to be

as concrete as possible and not sound like I’m insane.


And I’m not insane.

It’s obvious because you,

I’m a chemist.

So as a chemist, I grew into the world

understanding irreversibility.

Irreversibility is all I knew.

And when people start telling me

the universe is actually reversible, it’s a magic trick.

We can use time to do it.

So what I mean is that the second law

is really the magical.

But why does it need to be magical?

The universe is just asymmetric.

All I’m saying is the universe is asymmetric

in the state production and we can erase those states,

but we just have more computational power.

So what I’m saying is that the universe’s

deterministic horizon,

this is one of the reasons we can’t live in a simulation,

by the way, you can’t live in a simulation.

The irreversibility.

Yeah, yeah, so basically every time

you try and simulate the universe,

in this, you know, I live in a simulation,

the universe is expanded in states.

You’re like, oh damn it,

I need to make my computer bigger again.

And every time you try and contain the universe

in the computation,

because it’s got bigger in number of states.

And so I’m saying the fact the universe has novelty in it

is going to turn out experimentally to be proof

that time, as I’ve labeled it, is fundamental

and exists as a physical thing that creates space.

Okay, so if you can prove that novelty

is always being created,

you’re saying that it’s possible to also then prove

that it’s always expanding in the state space.

Those are things that have to be proven.

That’s what we’re working on experiments for, yeah.

And you’re trying to, like by looking at the sliver

of reality, show that there’s always novelty being generated.

Yeah, because if we go live in a universe

that conventional physicists would live in,

it’s a big lookup table of stuff and everything exists.

I want to prove that that book doesn’t exist,

it’s continuously being added pages on.

So all I’m saying, if the universe is a book,

we started, the universe at the beginning only had no pages

and had one page, another page, another page.

Whereas a physicist would now say all the pages exist

and we could in principle access them.

I’m saying that is fundamentally incorrect.

Do you know what’s written in this book?

The free will question.

Is there room for free will in this view

of the universe is generating novelty

and getting greater and greater assembly structures built?





Next question.

Why, what’s the source of free will in this?

Well, I think it depends on what you mean by free will, but.

Yeah, well, please, there’s a lot.

I think what I’m interested in

as far as the phenomena of free will is,

do we have individual autonomy and agency?

And when I do things, is it really me

or is it my atoms that did it?

And that’s the part that’s interesting to me.

I guess there’s also the determinism versus randomness part.

But the way I think about it is like each of us

are like a thread or like an assembly space

through this giant possibility space.

And it’s like we’re moving on our own trajectory

through that space and that is defined by our history.

So we’re sort of causally contingent on our past,

but also because of the sort of intersection

of novelty generation, it’s not completely predetermined

by the past.

And so then you have the causal control

of the determinism part that you are your causal history

and there’s some determinism from that past,

but there’s also room for creativity.

And I think it’s actually necessary

that something like free will exists

if the universe is gonna be as creative as possible.

Because if I were all intelligent being,

inventing a universe, and I wanted it to have

maximum number of interesting things happen,

again, we should come up with the metric of interesting,

but generating, yes, I know,

generating maximal possibilities,

then I would want the agents to have free will

because it means that they’re more individual,

like each entity actually is a different causal force

in the universe.

And it’s intrinsic and local property of that system.

There’s a greater number of distributed agents?

Like are you always creating more and more individuality?

Kind of.

I would say you’re creating more causal power, but.

So causal power, the word consciousness,

is the causal power somehow correlated with consciousness?

I mean, that’s why I have this conception

of consciousness being related to imagination,

because the more that we can imagine can happen

and the more counterfactual possibilities you have in mind,

the more you can actually implement.

And somehow free will is also at the intersection

of the counterfactual becoming the actual.

So can you elaborate on that a little bit,

that consciousness is imagination?

I don’t know exactly how to articulate it,

and I’m sure people will aim at certain things I’m saying,

but I think the language is really imprecise,

so I’m not the best way to.

It’s really interesting, like what is imagination

and what role does it play in the human experience,

in the experience of any agent?

Yes, I love imagination.

I think it’s like the most amazing thing we do.

But I guess one way I would think about it is,

we talked about the transition to life

being the universe acquiring memory,

and life does something really interesting.

You just think about biology generally.

It remembers states of the past to adapt

to things that happen in the future.

So the longer life has evolved on this planet,

the deeper that past is, the more memory we have,

the more kinds of organisms and things.

But what human level intelligence has done

is quite different.

It’s not just that we remember states

that the universe has existed in before,

it’s that we can imagine ones that have never existed,

and we can actually make them come into existence.

And I think that’s the most unique feature

about the transition to whatever we are

from what life on this planet has been doing

for the last four billion years.

And I think it’s deeply related to the phenomenon

we call consciousness.

Yeah, I was gonna, I mean, just agree with that.

I think that consciousness is the ability

to generate those counterfactuals.

Now, whether you can say, you know,

are there degrees of consciousness?

I mean, I’m sorry, panpsychist,

but electrons don’t have counterfactuals,

although they do have some kind of,

they are able to search a space and pathways.

But I think that there is a very concrete,

or concrete, there’s a very specific property

that humans have, and I don’t know if it’s unique to humans.

I mean, maybe dogs can do it, and birds can do it, right?

And where they are basically solving a problem,

because consciousness was invented,

or this abstraction was invented by evolution

for a specific reason.

And so look, one of the reasons why I came to the conclusion

that time was fundamental was actually because

Sarah and I had a completely different.

The most heated debate on Skype chat ever.

No, no, no, we had to, I was like.

Stop it.

No, no, it goes back to the free will thing.

So I think that, although I’ve changed my view a bit,

because there’s some really interesting physicists

out there who talks about how the measurement problem

in Newtonian space, but I don’t want to go there just now,

because I think I’ll mess it up.

But briefly, I could not see how the universe,

how we can have free will.

And I mean, this is really boring,

because this is like, this is a well trodden path,

but I mean, not so boring.

I suppose it’s kind of, we just want to be precise.

If the universe is deterministic,

how can we have free will, right?

So Sarah’s a physicist.

I think she, not believe, can show that most of the laws

we have are deterministic to some degree.

Quantum mechanics onto Newtonian stuff.

And yet there’s Sarah telling me she believes in free will.

And I’m like, your belief system’s broken here, right?

Because you’re demanding free will

in a deterministic universe.

And then I realized that I agreed with her

that I do think that free will is a thing

because we are able to search for novelty.

And then that’s where I came to the conclusion

that time, the universe is expanding in terms of novelty.

And it goes back to that Dan Dennett essay

that we’re talking about, the free will inflation.

Free will, so you have, so the past,

it did not exist in the past.

The past exists in the present.

What I mean is like, you are the, there was no past.

There is only present.

So that means you are the sum total.

Everything that’s occurs in the past

is manifestly here in the present.

And then you have this little echo state

in your consciousness because you’re able to,

you’re able to imagine something without actualization.

But the fact you imagine it, that occurs in electrons

and potassium ion flows in your neural network

in your brain.

Maybe consciousness is just the present.

So somehow you imagine that, and then by imagining,

oh, that’s good, yeah.

I’m gonna make a robot to do this thing and program it.

And then you physically then go and do it.

So that then changes the future, sorry.

What’s imagination?

Does it require the past?

Does it require the future?

Does it require memory?

Does it?

It’s imagination.

Does it only exist in the moment?

So imagination is, well, yeah, probably

it’s an instantaneous readout of what’s going on.

You can maybe, your subconscious brain

has been generating all the bits for it,

but no, imagination occurs when you,

in your game engine, you remember the past

and you integrate sensory to the present

and you try and work out what you want to do in the future.

And then you go and make that happen.

So the imagination is this, it’s like, imagine,

asking what imagination is about, asking what surfing is.

You can see, you can surfboard, surfer, wave coming in.

When you’re on that wave and you’re surfing,

that’s where the imagination is.

I think imagination is just accessing things

that aren’t the present moment in the present moment.

So like I’m sitting here and I’m looking at the table

and I can imagine the river and things or whatever it was.

And so it seems to be that it’s like,

it’s our ability to access things that aren’t present.

So conjure up worlds, some of them might be akin

to something that happened to you recently.

Right, but they don’t have to be things

that actually happened in your past.

And I think this gets back to assembly theory.

Like the way I would think about imagination

from an assembly theoretic standpoint

is I’m a giant causal graph and I exist in a present moment

as a particular configuration of Sarah.

But there’s a lot of, I carry a lot of evolutionary baggage.

I have that whole causal history

and I can access parts of it.

Now, when you talk about getting to something

as complex as us, having as large assembly space as us,

there’s ways of, like, there’s a lot of things

in that causal graph that have ever actually never existed

in the past history of the universe

because like the universe got big enough

to contain the three of us in this room in time,

but not all the features of each one of us individually

have come into existence as physical objects

we would recognize as individual objects.

This goes back to your point

that we actually have to explain

why things actually even look like objects

and aren’t just a smear of mass.

And just on the free will and physics thing,

when you were talking, I was, I just wanna bring this up

because I think it’s a really interesting viewpoint

that Nicholas Jisen has that, you know,

like we wanna use the laws of physics

and then say you can’t have free will.

And his point is you have to have free will

in order to even choose to set up an experiment

to test the laws of physics.

So in some sense, free will should be more fundamental

than physics is because to even do science,

there’s some assumption that the agents have free will.

And I always thought it was really perplexing

that, you know, physics wants to remove agency

because the idea that I could do an experiment here

on this part of earth and then I can move somewhere else

and prepare an identically, you know,

identically prepared experiment, run an experiment again,

seems to imply something about the structure of our universe

that is not encoded in the laws

that we’re testing in those experiments.

So this kind of dream of physics

that you can do multiple experiments, different locations

and then validate each other,

you’re saying that’s an illusion?

No, I’m saying that requires decision making

and free will to be a real thing, I think.

Like I think the fact that we can do science

is not arbitrary.

And I think people, you know,

the standard canon in physics would be,

well, you could trace all of that back

to the initial condition of the universe,

but the whole point of science is

I can imagine doing the experiment and I can do it

and then I can do it again and again and again

all over the planet.

To you, imagination is somehow fundamentally

generative of novelty.


So it’s not like the universe could have predicted

the things you imagined.

Imagination, so coming back to novelty,

I think novelty can exist outside of imagination,

but it supercharges it, it’s another transition, I think.

I mean, I would say,

I mean, this may be a boring statement,

but I would say the fact that, sorry?

I’m not sure, these are hard questions.

Yeah, I mean, I think the fact that objects exist

is yet another proof that time is fundamental

and novelty exists, right?

Because I think, again, if you ask a physicist

to write down in their infinite Bible of the universe,

let’s call it the Bible, the Mac, you know.

The book.

Yeah, well, I mean, the mathematical universe,

whether you’re Max Tegmark or Sean Carroll

or Frank Wilczek.

Or Stephen Wolfram, okay?

I like that book.

Yeah, I love it too.

It’s lots of pretty pictures.

It’s really interesting that they cope with the enormity

of the universe by saying,

well, it’s all there, mathematics, it all exists, right?

And I would say that that’s why I’m excited

about the future of the universe

because although it is somehow dependent upon the past,

it is not constrained just by the past,

which is kind of mad.

Yeah, that’s what free will is.

It’s not constrained by the past.

It’s dependent on the past.

This moment, it’s not just dependent,

this moment is the past,

and yet it has the capacity

to generate a totally unpredictable future.

I mean, the other thing I would say

that’s super important for human beings, right?

Human beings have actually very little causal control

in the future.

I realized this the other week.

Oh, the immediate future immediately.

Yeah, yeah, so what happened,

so this is what I think it is.

The way, by reinterpreting your past,

I mean, talk about from a kind of cognitive,

psychological cognitive point of view,

by reinterpreting your past in your current mind,

you could actually help you shape your future again.

So you have much more freedom to interpret your past,

to act in the present, to change your future,

than you do to change your future.

It may sound weird, so I’m saying to everybody,

imagine your past, think about your past,

reinterpret your past in the nicest way you can,

then imagine what you can do next,

or imagine your past in a more negative way

and what you do next,

and look at those two counterfactuals, they’re different.

Yeah, it’s fascinating.

I mean, Daniel Kahneman talks about this

that most of our life is lived in our memories.

It’s interesting, because you can essentially,

in imagination, choose the life you live.

So maybe free will exists in imagination.

Choices are made in your imagination,

and that results in you basically able to control

how the future unrolls, because you’re like,

imagining, like reinterpreting constantly

the things that happen to you.

Exactly, so if you want to increase

your amount of free will, those people that have,

I don’t think everyone has equal amounts of agency,

because of our sad constraints,

where they’re happenstance, health, economic,

born in a certain place, right?

But those of us that have the ability to go back

and reinterpret our past and use that to change the future

are the ones that exert most agency in the present,

and I want to achieve higher degrees of agency

and enable everyone else to do that as well,

to have more fun in the universe.

Then we’ll hit that peak, the maximum fun point.

I don’t think there’s ever gonna be a maximum,

I think the wonderful thing about the future

is there’s always gonna be more fun.

Yeah, you, I think, again, going back to Twitter,

I think you, Lee, tweeted something

about being a life maximalist,

that you want to maximize the number of life,

the amount of life in the universe, so,

and that’s the more general version of that goal,

is to maximize the amount of fun in the universe,

because life is a subset of fun, there are all kinds of,

I suppose they’re either correlated or exactly equal,

I don’t know.

Anyway, speaking of fun,

let me ask you about alien sightings.

So there’s been quite a bit of UFO sightings

and all that kind of stuff.

What do you think would be the first time

when humans sight aliens, see aliens,

in a sort of unquestionable way?

This extremely strong and arguable way

we’ve made contact with aliens.

Sarah, what would it look like?

Obviously, the space of possibilities is huge here,

but if you were to kind of look into the future,

what would that look like?

Would it be inklings of UFOs here and there

that slowly unravel a mystery?

Or would it be like an obvious, overwhelming signal?

So I think we have an obsession

with making contact with events.

So what do I mean by that is,

like people have a UFO sighting, they make contact.

And I always think, you know,

what’s interesting to me about the UFO narratives right now

is not that I have a disbelief

about what people are experiencing or feeling,

but like the discussion right now

is sort of at the level of modern mythology.

Aliens are our mythos in modern culture.

And when you treat it like that,

then I wanna think about when do things

that we traditionally only regularize through mythology

actually become things that become standard knowledge?

So, you know, like it used to be, you know,

variations in the climate were described

by some kind of gods or something.

And now it’s like, you know,

our technology picks up an anomaly

or someone sees something, we say it’s aliens.

And I think the real thing is it’s not contact with events,

but like first contact is actually contact

with knowledge of the phenomenon or the explanation.

And so this is very subtle and very abstract,

but when does it become something

that we actually understand what it is

that we’re talking about?

That’s first contact.

It’s not.

Would you make the myth,

would you give credit to the myth,

the mythology as first contact?

Cause you might.

I think, yes.

I think it’s the rudimentary

that we have some understanding that there’s a phenomenon

that we have to understand and regularize.

So I think.

Right, to understand that there is weather.


You have to construct a pathology around that weather.

Yes, yes.

It’s something that’s controllable.


I see mythology basically as like baby knowledge.

Right, right.

It could be that, you know,

although there’s lots of alien sight,

so called alien sightings, right?

So there is a number of things you can do.

You could just dismiss them and say they’re not true.

They’re kind of made up.

Or you say, well, there’s something interesting here, right?

We keep seeing a commonality, right?

We see the same phenomenon again and again and again.

Also, there’s this interesting thing

about human imagination.

Even if they are, let’s not say made up,

but misappropriated kind of other inputs,

the fact that human consciousness

is capable of imagining a contact with aliens.

Does that not tell us about something

about where we are in our position,

in our culture, in our technology?

It tells us about where in time we are.

Could it be that we’re making contact with,

let’s say that, so let’s say,

let’s take the most miserable version.

There are no aliens in the universe.

Life is only on Earth.

That then, the interpretation of that

is we’re desperate to kind of understand

why we’re the only life in the universe, right?

The other one is the other most extreme

is that aliens are visiting all the time.

We’re just not able to capture them coherently,

or there’s a big conspiracy and there’s Area 51

and there are lizards everywhere and there’s that.

Or I’m kind of in favor of the idea

that maybe humanity is waking up to the idea

that we aren’t alone in the universe

and we’re just running the simulation

and we’re seeing some evidence.

We don’t know what life is yet.

We do have some anomalies out there.

We can’t explain everything.

And over time, you know, we will start to unpack that.

One very plausible thing we might do,

which might be boring for the average alien observer

or believes that aliens are,

as in intelligent aliens are visiting Earth,

it could be that we might go to the outer solar system

and find a new type of life

that has completely new chemistry,

bring these cells back to Earth,

where you could say in my hand,

on Earth, here’s RNA, DNA, and proteins,

and look, cells self replicate.

From Titan, we got this new set of molecules,

new set of cells, and we feed it stuff and it grows.

That for me, if we were able to do that,

which would be like the most,

that would be my UFO sign.

That’s a good test.

So you feed it and it grows.


We’ve made, so not until you know how to feed the thing,

it grows somehow.

We can make a comic book, you know,

the tiger that came for tea, the alien that came for tea.

What would you say is between the two of you

is the biggest disagreement about alien life out there?

Is it from the basic framework of thinking about

what is life to maybe what aliens look like

to alien civilizations, to UFO sightings?

What would you think?

So I would say the biggest one is that

the emergence of life does not have to be,

that it can’t just happen once on the planet,

that it could be two or more life forms

present on the planet at once.

And I think Sarah doesn’t agree with that.

I think that’s like logically inconsistent.

That’s really polite.

You’re saying it’s nonsense.

Because you think that, yeah.

So likely is that, so the idea that,

what does it look like?

Let’s imagine two alien civilizations

coexisting on a planet.

What’s that look like exactly?

So I would say, I think I’ve got to get around your argument.

Yeah, let’s say that on this planet,

there’s just like, there’s lots of available chemistry

and one life form gets some emerges based on carbon

and interacts and there’s an ecosystem based on carbon.

And there’s an orthogonal, and so it’s planetary phenomena,

which is what you, I think, right?

But there’s also one that carries on silicon.

And because there’s enough energy and there’s enough stuff,

that these life forms might not actually

necessarily compete evolutionarily.

Yeah, but they would have to not interact at all

because they’re going to be co constructing

each other’s causal chains.

I think that’s what you just got me, yeah.

So there’s no overlap in terms of their causal chains

or a very limited overlap.

Yeah, so I think the only way I can get away with that

is to say, right, life can emerge on a planet underneath.

And, okay.

The lizard people under the crust of the earth.

I think, I think, I think we, let’s go to,

I think, but look, as you can see, we disagree.

So, and I think Sarah actually has convinced me

because of the life is a planetary,

the emergence of life is a planetary phenomena.

And actually, because of the way evolution selection works,

then nothing occurs in isolation.

The causal chains interact.

So there is a common, there’s a consensus model

for life on the earth.

But you don’t think you can place aliens from elsewhere

onto the, can’t you just place multiple alien civilizations

on one planet?

Right, but I think, so you can take two origin life events

that were independent and co mingle them.

But I don’t think when you’re talking about, when you,

when you look at the interaction of that structure,

it’s like the same idea as like an experiment

being an example of life, right?

That’s a really abstract and subtle concept.

And I guess what I’m saying is life is information

propagating through matter.

So once you start having things interacting,

they in some sense co mingle and they become part

of the same chain.

So the co mingling starts quickly, proceeds,

we proceed to co mingle quickly.

Right, right, so you could say, so the question is then,

the more interesting question is,

are there two distinct origins events?

And I still think that there’s reasons that

on a single planet, you would have one origins event

because of the timescales of cycling of geochemistry

on a planet and also the fact that I don’t think

that the origin of life happens in a pool

and like radiates outward through evolutionary processes.

I think it’s a multi scale phenomenon happens

at the level of individual molecules interacting,

collections of molecules interacting

and entire planetary scale cycles.

So life as we know it has always been multi scale

and there’s brilliant examples of individual mutations

at the genome level changing global climate, right?

So there’s a tight coupling between things that happen

at the largest scale, our planetary scale

and the smallest scale that life mediates.

But it still might be difficult within something

you would call as a single alien civilization.

You know, there’s species and stuff.

And they might not be able to communicate.

But you’re asking about life, not species, right?

What’s the difference between one living civilization?

This is almost like a category question.

Versus species, because it can be very different.

Because there’s like literally islands

that you can evolve different kinds of turtles and stuff.

And they can.

So I guess what I’m saying is weird.

If you look at the structure

of two interacting living things, populations,

and you look in their past

and they have independent origins for their causal chain,

then you would say one was alien.

You know, they have different independent origins events.

But if you look at their future

by virtue of the fact they’re interacting,

their causal chains have become commingled.

So then in the future, they are not independent.

So that’s why you would even define them as alien.

So the structure across time is two examples of life

become one example of life

because life is the entire structure across time.

Right, but there could be a lot of variation within.

Yeah, so the question we’re all interested in

is how many independent origins

of a complexifying causal chain are there in the universe?

See, but the idea of origin is easy for you to define?

Because like, when the species split

in the evolutionary process,

and you get like a dolphin versus a human

or a Neanderthal versus Homo sapiens, isn’t there?

Let me make a distinction here quickly.

So I think, sorry to interrupt.

What we’re saying, I mean, Sarah won that argument

because I think she’s right,

that once the causal chains interact and going forward,

so we’re talking about a number of things.

Let’s go all the way back before origin of life.

Origin of life.

On Earth.

Chemistry emerges, so there’s all these,

I would say there’s probably mechanistically,

the chemistry is desperately trying to find any way

to get replicators.

The ribosome kind of was really rubbish at the beginning

and they just competed, competed, competed,

and you got better and better ribosome.

Suddenly that was a technology.

The ribosome is the technology that, boom,

allowed evolution to start.

So what I was trying to, why I interrupted you,

is say that once evolution has started

using that technology, then you can speciate.

And I was trying to, and I think what Sarah said

was, convinced me of, because I was like, no,

we’re gonna have lots of different chemistry,

shadow biosphere on Earth, and she’s like, no, no, no.

You have to have this, you have to get to this

minimum evolutionary machine.

And then when that occurs, speciation occurs.

Exactly what it’s like, dolphins, humans,

everything on Earth.

But when you’re looking at aliens or alien life,

there’s not gonna be two different types of chemistry

because they compete and interact and cooperate

because the causal chains overlap.

One might kill the other, one might combine with the other,

and then you go on and then you have this average.

And sure, there might be respeciation.

It might have two types of emerging chemistry.

It almost looks like the origin of life on Earth

required two different prelife forms,

the peptide world and the RNA world.

Somehow they got together, and by combining,

you got the ribosome.

And that was the minimum competent entity for evolution.

And would all alien civilizations

have an evolutionary process on a planet?

So it’s almost the definition of life.

To create all those memories, you have to have something.

Things have to change in time.

But there has to be selection.

That’s like an efficient, there’s no other way to do it.

No, well, never say never, because soon as I say that.

That’s the part that depresses me, though,

going back to the earlier discussion on violence

and things, and I don’t know where,

somebody was tweeting about this recently,

but how much death had to die.

Maybe it was you.


Ah, yeah.

So, yeah, sorry.

We’re talking about life.


And I guess a lot of murder had to occur.

Right, so selection means things had to be weeded out,

right, so.

Well, we can celebrate that.

Death makes way for a tulip.

Yeah, I mean, and also one of the most interesting features

of major extinction events in the history of our planet

is how much novelty emerged immediately after, right?

So, and of course, a lot of people make arguments

we wouldn’t be here if the dinosaurs didn’t go extinct.

So, in some ways, we can attribute our existence

to all of that.

But I guess I was just wondering and sort of like,

if I was gonna build a universe myself

in the most optimistic way, would I retain that feature?

But it does seem to be a universe.

I think you have to.

I mean, I think we’re probably being

over anthropomorphizing.

I remember watching the blue,

I think it was the blue planet,

David Attenborough was showing these seals

and because of climate change,

some seals were falling off a cliff.

And how tragic that was.

I was like, I’m saying my son, that’s pretty cool.

Look at those ones down there.

They’ve obviously got some kind of mutations,

some and they’re not doing that Darth thing.

And so that poor gene will be weeded out.

Of course, at the individual level, it looks tragic.

And of course, as human beings have the ability

to abstract and we empathize,

we don’t wanna cause suffering on other human beings

and we should retain that.

But we shouldn’t look back in time

and say, how many butterflies had to die?

I remember making this,

how many, if you think about the caterpillar

become the chrysalis and then the butterfly getting out,

how many, if that suffering, we call it suffering,

if that process of pruning had not occurred,

we have no butterflies.

So none of the butterfly beauty in the world

without all that pruning.

So pruning is required,

but we shouldn’t anthropomorphize

and feel sorry for the biological entities,

because that seems to be a backwards way of looking at it.

What we should do is project forward

and maybe think about what values we have across our species

and our ecosystem and our fellow human beings.

You know, now that we know that animals suffer

at some level, think about humane farming.

When we find that plants can, in fact,

are conscious and can think and have pain,

then we’ll do humane gardening.

Until that point, we won’t do it, right?

I like this.

Famous chemist endorses the majestic nature of murder.

That’s the title.

I didn’t say that, but any case.

Well, I just inserted it.

I have a hard time with it, though.

I think the way you put it is kind of…

But it’s the reality of, it is beautiful.

You know, there’s an Instagram account

called natureismetal,

and I keep following it and unfollowing it

because I can’t handle it for prolonged periods of time.

We evolve together, you die alone.


We evolve together, but you die alone.

We live alone, too.

It’s the Gatsby thing.

I don’t know.

We evolve together.

Where’s the together?

The together is the murder and the sex, sex and murder.

My romantic vision of it is to try to make me happy, Sarah,

instead of sad Sarah.

I talk in third person when I think very abstractly.

Sorry, is, you know, like this whole,

like certain things can coexist,

so the universe is trying to maximize existence,

but there’s some things

that just aren’t the most productive trajectory together,

but it doesn’t mean that they don’t exist

on another timeline or another chain somewhere else.

Like, and maybe you would call that, like,

then some kind of multiverse or things,

but what am I saying?

I think you can’t.

I just, you can’t go down a level.

I’m just making stuff up.

No, you’re not.

It makes me feel better.

I don’t understand.

Is it logical?

And we need, we need.

No, I know, I know.

Yeah, if you look at bacteria, if you look at virus,

I mean, just the number of organisms that are constantly,

like looking at bacteria, they’re just dying nonstop.

It’s like a slaughter.

Right, right.

Well, and this goes back to the conversation about God.

I mean, like, there’s the whole thing about, like,

why is the universe unable to suffering?

Individuals don’t exist, right?

Individuals, so for this, I think,

if you think about life as an entity on Earth, right?

Let’s just, let’s just go back a second.

I mean, I like to, I’ll be ludicrous for a second.

I don’t exist.

You don’t exist, right?

But you, but the actions you do,

the product of evolution exists, right?

The objects you create exist quantitatively

in the real world.

If you then understand life on Earth or alien life

or any life in the universe as this integrated entity

where you need, you need cells in your body to die.

Otherwise you’d just get really big

and you wouldn’t be able to walk around, right?

So, you know, you do.

Yeah, yeah, yeah, yeah.

So, so I think.

It’s the patterns that persist, not the physical things.

And of course, we, you know, we have, we have,

we place immense values on fellow human beings

and I’m, majestic professor does like

other individual human beings.

Now you’re talking in third person too.

I know, it happens, right?

So death, would you say, I mean,

because you said evolution is a fundamental part of life.

So death is a fundamental part of life.


It might, right now, it might not be in the future.

We might hack some aspects of death

because, and we’ll evolve in different ways.

But isn’t there, I think Sarah mentioned

like this life density.

Is it, can’t that become a problem?

Like too much, too much bureaucracy,

too much baggage builds up.

Like you need to keep erasing stuff.

I think it’s okay that we dissipate.

Like, I don’t think of it like, like, I mean.

Dissipate, yes.

No, but I mean, like, like we’re so fixated

on ourselves as individuals and agents.

And we were talking about this last night,

actually over dinner, but like,

you know, an individual persists

for a certain amount of time,

but what you want to do,

like if you’re really concerned with immortality

is not to live indefinitely as an individual,

but maximize your causal impact.

So like, what are the traces of you that are left?

And you’re still a real, I always think of Einstein,

like for a period of time, he was a real physical thing

we would identify as a human.

And now we just see echoes of that human

in all of the ways that we talk about his,

you know, causal impact are frankly great

is another great example.

How many Easter eggs could you leave in the future?

It’s like, oh, I got you.

So I guess the question is,

how much do you want to control the localization

of certain features of say a prop,

a packet of propagating information we might call person

and keep them localized

to one individual physical structure?

Or do you want to, you know,

is there a time when that just becomes a dissipated feature

of the society that it once existed in?

And I’m okay with the dissipated feature

because I just think that makes more room

for more creativity in the future.

So you mentioned engineering life in the lab.

Let me take you to computer science world.

What about robots?

So is it possible to engineer,

so you’re really talking about like engineering life

at the chemistry level,

but do you think it’s possible to engineer a life

at the like humanoid level, at the dog level?

Like, or is that, like at which level can we instill

the magic of life into inanimate stuff?

No, I think you could do it at every level.

I just think that we’re particularly interested

in chemistry because it’s the origin of life transition

that presumably, or at least that’s how I feel about it,

it’s going to give you the most interesting

or deepest insights into the physics.

But presumably everything that we do and build

is an example of life.

And the question is just how much do you want to take

from things that we have now and put them into,

like examples of life and copy them into machines?

I saw that there was this tweet again.

I think you were at the Mars conference

and you were hanging out with a humanoid robot.

Yes, that was a fun time.

Making lots of new friends at Mars 2020.

Did you guys color match ahead of time with the robot

or did that accidentally happen?

Accidentally, I went up and I wanted to say hi.

Torpoise, would that be the correct name for the color?

I think so.

We didn’t color coordinate our outfits.

Well, you didn’t, maybe the robot did.

The robot probably did, much more stylish.

So for people who are just listening,

there’s a picture of Sarah standing next

to a humanoid robot.

I guess you like them with a small head and perfect vision.

Actually, no, I just.

What did they perfectly, there’s a LIDAR.

No, I mean, I think I was just deeply interested


What was, sorry to interrupt, was it manual control?

Was it actually stabilizing itself?

Oh no, it was walking around.

Oh, nice.



It was pretty impressive.

I mean, actually there’s some videos online

of Jeff Bezos walking with one of those

across the lawn nearby there.

This is great.



I wasn’t invited.

Yeah, but there you go.


That’s incredible, isn’t it?


See, you look at the walking robot.

Where did the idea for walking come from?

It was invented by evolution, right?

And us as human beings, able to conceptualize and design

and engineer the causal change.

So that robot is evidence of life.

And so I think what’s going to happen is there’s the,

we want to find where the spark comes from mechanistically.

How can you literally go from sand to cells?

So that’s the first transition that I think,

there are a number of problems we want to do.

Make life in the lab.


Then we want to make life in the lab

and want to suddenly start to make intelligent life

or life that can solve, start to solve abstract problems.

And then we want to make life that is conscious.


In that order?

I think it has to happen in that order.

Getting towards this artificial general intelligence.

I think that artificial general intelligence

can’t exist in a vacuum.

It has to have a causal change

all the way back to Luca, right?


And so the question I think,

I really like the question is to say,

what are we, how is,

how is our pursuit of more and more lifelike?

I know you want to, you like your robots.

You want to project into them.

You want to interact with them.

I think you would want,

if you have a robot dog and the robot dog

does everything expected of a normal dog

and you can’t tell the difference,

you’re not really going to ask the question anymore.

If it’s a real dog or not,

or you’ve got a personality, you’re interacting with it.

And so I think what would be interesting

would be to kind of understand

the computational architecture, how that evolves.

Cause you could then, you know,

teleport the personality from one object to the other

and say, right, does it act the same?

And I think that as we go along,

we’re going to get better and better

at integrating our consciousness into machines.

Well, let me ask you that question just to link on it.

I would, I would call that a living conscious thing

potentially, I as a human allegedly,

but would you, as a person trying to define life,

if you pass the Turing test, are you a life form?

One of the reasons I walked up to the robot

was because I wanted to meet the robot, right?

So I, it felt like I was,

and I abase a lot of my interaction with reality

on emotion and feeling,

but like how do you feel about an interaction?

And I always love your point about like,

is it enough to have that shared experience

with a robot, right?

So walking up to it,

does it feel like you’re interacting with a living thing?

And it did to an extent, but in some degrees,

it feels like you’re interacting with a baby living thing.

So I think our relationship with technology

and particularly robots we build is really interesting

because basically they exist as objects in our future

in some sense, like we’re a much older evolutionary lineage

than robots are, but we’re all part of the same causal chain

and presumably, you know, they’re kind of in their infancy.

So it’s almost like you’re looking at the future of life

when you’re looking at them,

but it hasn’t really become life in a full manifestation

of whatever it is that they’re gonna become.

And, you know, the example of the walking robot

was super interesting, but they also had a dolphin

that they put in the pool at the cocktail party at Mars,

and it looked just like a real dolphin swimming in the pool.

And, you know, it’s in this kind of uncanny valley

because, and I was having this conversation

with a gentleman named Mutu who was super perceptive,

but he was basically saying like,

it made him feel really uncomfortable.

And I think.

The dolphin.

Yeah, and I think a lot of people would have that response.

And I guess my point about it is it is kind of interesting

because you’re basically trying to make a thing

that you think is nonliving mimic a living thing.

And so the thought experiment I would wanna run in that case

is imagine we replaced every living thing on earth

with a robot equivalent, like all the dolphins and things.

And in some sense, then you’re making,

if you think that the robots aren’t experiencing reality,

for example, in the way that

a biologically evolved thing would,

you’re basically making the philosophical zombie argument

become real and basically building reality into a simulation

because you’ve made everything quote unquote fake

in some sense.

You’ve replaced everything with a physical simulation of it.

So as opposed to being excited by the possibility

of creating something new,

you’re terrified of humans being replaced.

I was just trying to run like what would be

the absolute thought experiment.

But I don’t think that scenario would actually play out.

I guess what I think is weird for why we feel

this kind of uncanny valley interacting

with something like the robot dolphin

is we’re looking at an object we know is kind of

in the future in the sense of like

if everything’s ordered in time,

but it’s borrowing from a structure

that we have common history with,

and it’s basically copying in a kind of superficial way

things from one part of the causal chain to another.


Well, that’s a video of…

Everybody believed it was real.

They look so real.

And obviously the technology was developed for movies, so.

Well, I think we’re confusing our emotional response

and understanding the causal chain

of how we got there, right?

Because the philosophical dot zombie argument

thinks about objects just appearing, right,

that you’re facsimile in some way,

whereas there is the causal, the chain of events

that caused the dolphin to be built went for a human being.

Yeah, would a philosophical zombie

still have a high assembly index?


Because it can’t be, philosophical zombies can’t,

like Boltzmann brains, just can’t appear out of nowhere.

Well, I guess my question would be in that scenario

where you built all the robots

and replaced everything on Earth with robots,

would the biosphere be as creative

under that scenario or not?

Yeah, that’s a good question.

Are there quantitative differences

you would notice over time?

And it’s not obvious either way, right?

It’s not obvious right now,

because we don’t really, we don’t understand,

we haven’t built into machines how we work.

So that’s, I think, one of the big missing things

that we’re both looking for, right?

This is a robot, it’s a cute robot.

But the point, Sarah, is that the biosphere

won’t be as creative if you did it right now.

No, of course, I think that’s why people don’t like it.

But in the future, we will be able to solve the problem

of origin of life, intelligence, and consciousness

because they exist in physical substrates.

We just don’t understand enough

about the material substrate and the causal chain.

But I’m very confident we will get to an AGI,

but it won’t be what people think.

It won’t be, solution won’t be a, we’ll get fooled a lot.

And so GPT3 is getting better at fooling us

and GPT153 might really fool us,

but it won’t have the magic we’re looking for.

It won’t be a creative,

but it will help us understand the differences between.

Really though, because isn’t that what love is, being fooled?

Like what, why are you not giving much value

to the emotional connection with objects,

with robots, with humans?

Emotion is a thing which happens

when your expectation function is dashed

and something else happens, right?

I mean, that’s what emotion is.

Is that what love is too?


You were expecting one thing and something else happened.


I don’t know.

I don’t think that’s true either.

Well, what is it then?

I think, no, emotion, look, I’m sorry, emotion is that,

but there’s no more.

No, I think love is just fulfilling your purpose.

No, but, okay, I mean, look, look.

Like whatever that means.

That’s the opposite of what’s gonna happen.

I mean, really?

So, okay.

I think the happiest is like when you’re doing.

All right, all right, all right.

Let me go back.

If you want me to define.

Follow your bliss.

Let me define love quickly.

Okay, go for it.

In terms of assembly space, right?



I didn’t think I’d be doing this today.

I can’t wait till Assembly Theory 101 is taught

and the second lecture is Assembly Theory of Love.

No, no, but look, actually, but look, but.

It’s being surprised.

The expectation is being broken.

I’m just, I’m not.

No, go for it.

I wanna hear you.

I’m not an emotional being, but I would say,

so let’s talk, so we’ll talk about emotion a bit,

but love is more complex.

Love is a very complex set of emotions together

and logical stuff, but if you’ve got this thing,

this person that’s on this causal chain

that has this empathy for this other thing,

love is being able to project ahead in your assembly space

and work out what you’re,

the person you’re in love with has a need for

and to do that for them without selflessly, right?

Cause you can project ahead what they’re gonna need

and they are there and maybe you can see

someone who’s gonna fall over

and you catch them before they fall over

or maybe you can anticipate that someone’s gonna be hungry

and without helping you, you just help them.

That’s what love is.

That just sounds like empathy.

But it’s more complex than that, right?

It’s more complex.

It’s more about not just empathy, it’s understanding.

It’s about kind of sharing that experience.

That’s an expression of love though.

That’s not what it’s like to feel love.

Like feeling love is like,

I think it’s like when you’re aligned with things

that you feel like are your purpose

or your reason for existing.

So if you have those feelings towards a robot,

why is that robot?

I mean, cause you said like the AGI,

we’ll build an AGI,

but there’ll be a fundamental difference in AGI.

I don’t think we’ll build it.

It’s gonna emerge from our technology.

I think you guys all argue the same thing.

I just said that GPT,

we do not correctly capture the causal chain that we have.

Within GPT.

Yeah, within AI.

But don’t you think it captures,

because GPT3 is fundamentally trained

on a corpus of knowledge,

like the internet.

Don’t you think it gets better and better and better

at capturing the memory of all the…

It will be better at fooling you.

And at some point you won’t care.

But when it comes to,

my guess, this is a quick,

this is what I was getting to right before we got,

I got in the love trap.

Love trap, yeah.

It was like Lee Cronin in the love trap.

Sounds like a good fan name.

SAD, okay, SAD, assembly space of SAD.

No, is that, so sure.

But I think there are other features

that allow that we pull on innovation

that allow us to do more than what we just see in GPT3.

So if you’re being fooled there.

So I think what I mean is human beings have this ability

to be surprising and creative.

Whereas is it Dali, this thing,

or if you take GPT3 is not gonna create a new verb.

Shakespeare created new verbs.

You’re like, wow.

And that required Shakespeare to think outside of language

in a different domain.

So I think having that connections across multiple domains

is what you need for AGI.

Yeah, but I don’t know if you need,

I don’t know if there’s any limitations to GPT

and not being able to be across domain.

The number one problem is,

it’s instantiated in a resource limited substrate

and that we don’t, in silicon.

The architectures used for training, for learning

is about fooling, it’s not about understanding.

And I think that there is some understanding that we have

that is not yet symbolically representable.

Language, learning language and using language

seems to be fundamentally about fooling, not understanding.

Why do you use language exactly?

I might disagree with that quite fundamentally actually,

but I’m not sure I understand

how I make a coherent argument for that.

But my feeling is that there is comprehension in reality

in our consciousness below language.

And we use those for language,

for all sorts of expressions.

And we don’t yet understand that there’s a gap.

We will get there, but I’m saying,

wouldn’t it be interesting, it’s a bit like saying,

could I facsimile you or Sarah into a new human being?

And let’s just say, I could copy all your atoms

and the positions of all your atoms and electrons

into this other person, they would be you.

The answer is no.

And it’s quite easy to show using assembly theory

because actually the feature space that you have,

that graph, the only way to copy you

is to create you on that graph.

So everything that’s happened to you in your past,

we have to have a faithful record for.

If you want another copy of Lex,

you have to do the exact thing.

Want another copy of Sarah, want another copy of Lee.

The exact past has to be replicated.

Let me push back on that a little bit.

That’s maybe from an assembly theory perspective,

but I don’t think it’s that difficult

to recreate a version of me, like a clone,

that would make everybody exactly equally as happy.

They wouldn’t care which one.

And there’s two of me and then they get to pick which one

and they’ll kill either one, they’ll be fine.

As long as they’re forced to kill.

They’ll be fine.

But here’s what will happen is,

let’s say we make artificial Lex

and it was like, wow, so cool.

It looks the same interact.

Then there’ll be this battle of like,

right, we’re going to tell the difference.

We’re going to basically keep nudging Lex

and artificial Lex until we get novelty from one

and we’ll kill the other one.

And I think, thank God.

We’re not, novelty is a fuzzy concept.

That’s the whole problem of novelty.

So I will define novelty, it’s not fuzzy.

Novelty is the ability for you to create architectures

that are, or create an architecture.

So let’s say you’ve got a corpus of architectures known.

You can write down, you’ve got some distance measure.

And then I create a new one and the distance measure

so far away from what you’d expected.

There’s no linear algebra going to get there.

It’s like, that is creativity.

And we don’t know how to do that yet on any level.

Well, I was also thinking about like your argument

about free will, like you wouldn’t be able to know

it was, it doesn’t work instantaneously.

It’s not like a micro level thing,

but more a macro level thing over the scale of trajectories

or longer term decisions.

So if you think that the novelty manifests

over those longer timescales,

it might be the two Lexes diverge quite a bit

over certain timescales of their behavior.

But nobody would notice the difference.

They might not.

And the universe, the earth won’t notice the difference.

The universe won’t notice the difference.

The universe would notice the difference.

No, the universe doesn’t know about its novelty

that’s being generated.

That’s the whole point of novelty.

Yeah, but this is what selection is, right?

It’s like taking nearly equivalent ones

and then deciding like the universe selects, right?

So whatever selection is,

select some things to persist in time.

Yeah, it’s gonna select the artificial one

just because it likes that one better.

Well, you’re mixing up two arguments here.

So look, let’s go back a second.

What are you basing this argument on, Lex?

I’m just saying that I kind of don’t think,

cause at least said that it’s not possible.

Like if you copy every single molecule in a person’s body,

that’s not going to be the same person.

That they won’t have the same assembly index.

It won’t be the same person.

And I just don’t, I think copying, you can compress.

Not only do I disagree with that,

I just, I think you can even compress a person down

to some where you can fool the universe.

I’m saying, let me restate it.

It is not possible to copy somebody on,

because you, unless you copy the causal history.

Also, you can’t have two identical.

I mean, actually I really like the idea

that everything in the universe is unique.

So even if like there were two Lex’s.

I know you like that idea

cause you’re human and you think you’re unique.

Yeah, exactly.

But also I can make a logical argument for it

that even if we could copy all of your molecules

and all their positions, the other you would be there.

And you have a different position in space.

And the other thing.

You’re distinguishable.

Yeah, the other thing I was gonna add.

How unique are you?

Just by the position in space really.

Sure, but then how much does that light translation

of Lex sitting there affect the future?

I see, but wait a minute.

Is part of the definition of something being interesting

is how much it affects the future?


But let me come back.

Don’t you agree?

Do you disagree?

But let me come back one point quickly

that you were making.

Sure, I think I probably agree, yes.

There’s two Lex’s, right?

There’s a robot Lex that you just basically,

it is a charade.

It’s a facsimile.

It’s just coded to emulate you.

Are you robot Lex?

I wouldn’t know, right?

Let’s get that.

But let’s get that.

That’s the point.

It’s a very important point here

because he’s ducking and diving between this eye.

So if I facsimile you into a robot,

then your robot might be,

would be a representation of you now,

but fundamentally be boring

because you go and have other ideas.

If, however, you built an architecture

that itself is capable of generating novelty,

you would diverge in your causal chain

and you’d both be equally interesting to interact with.

We don’t know that mechanism.

All I’m trying to say is we don’t yet know that mechanism.

We do not know the mechanism that generates novelty.

And at the moment in our AIs, we are emulating,

we are not generating.

You don’t think we’re sneaking up on that?

No, no, there is no ghost in the machine.

And I want there to be one.

I want the same thing you want.

Sorry, I was.

I know you want that as a human

because everything you just said

makes you feel more special.

I want to be, no, no, no, screw my specialness.

I just want to be surprised.

If I.

You don’t think a robot can surprise you.

If I, if you can produce an algorithm

instantiated in a robot to surprise me,

I will, I will, I will, I will,

I will have one of those robots, it’ll be brilliant.

But they won’t, it won’t surprise me.

But why, why is it a problem

to think that humans are special?

Maybe it’s not the special, you’re right.

It’s the better than.


Because then you start to not recognize

the magic in other life forms

that you either have created or you have observed.

Because I just think there is magic

in legged robots moving about.

And they are full of surprises.


So this is.

And functionality.


So I’m a little.

I know why you like cellular automata, right?

But the specialness in your robot

comes from the roboticist that built it.


It’s part of the lineage.


And so that’s fine.

I’m happy with that.

That’s what I felt like looking at the standing robot

was I was looking at four billion years of evolution.

Yeah. Right.

If it wasn’t.

So I think I’m happy.

I mean, I’m happy we’re gonna coexist.

I’m just saying you’re gonna get more excitement.

There’s something missing

in our understanding of intelligence.

Intelligence isn’t just training.

The way the neural network is conceived right now is great

and it’s lovely and it’ll be better

and we will argue forever.

But you want to know, wouldn’t it be great if I said,

look, I know how to invent an architecture

and I can give it a soul.

And what I mean by a soul is some,

I know for real that there is internal reference.

Soon as I not fake internal reference.

And if we could generate that mechanism

for internal reference, that’s why our goal direct.

That’s why you have to develop a test for goal directness.

Get that goal directness.

You would love that robot more than the one

that’s just made to look like it does

because you’ll have more fun with it

because you better generate search, other problems,

get to more novelty.

Hell, you’d be able to fall in love with that robot for real,

but not the one that’s faking it.

What about fake it till you make it?

Well, I think a lot of people fall in love with fake humans.

It’s nice to fall in love with something

that’s full of novelty, yes.

I could imagine all kinds of robots

that I would want to have a close relationship with.

And I don’t mean like sexual, I mean like intimacy.

But I just don’t think that novelty generation

is such a special.

Okay, there’s like mathematical novelty

or something like that,

and then there’s just humans being surprised.

I think we’re easily surprised.

That’s fine, but that’s that.

But you don’t think that’s a good definition of novelty?

No, that’s good.

I’m happy to be surprised,

but not globally surprised because someone else,

but I really want, I was wondering why I’m a scientist.

I really want to be the first to be surprised about something

and the first thing in the universe

to create that novelty

and to know for sure that that novelty

has never occurred anywhere else.

That’s a real buzz, right?

Is there a way to really know that?

You have to have a really big lookup table.


Yeah, you’re never going to be know for sure, right?

That’s one of the hard things about being

a scientist searching for this type of novelty.

Maybe that’s why mathematics, mathematicians love discovery,

but actually they are creating.

And then when they create a new mathematical structure

that they can then, you can write code to work out

whether that structure exists before that.

That’s almost why I would love to have been a mathematician

from that regard to invent new math

that really I know pretty much for sure

does not exist anywhere else in the universe

because it’s so contingent.

Right, but this gets into like you said a few times

that I still really don’t understand

how you actually plan to do this,

to build an experiment that detects

how the universe is generating novelty

or that time is the mechanism.

So the problem that we all have,

which I think is what Lex is pushing against

is if I build the experiment,

you don’t know what you put into it.

So you don’t know what, like if you,

unless you can quantify everything you put in,

all of your agency, all the boundary conditions,

you don’t know if you somehow biased it in some way.

So is the novelty actually intrinsic to that experiment

or to that robot, or is it something you gave it,

but you didn’t realize you gave it?

It’s gonna be, it’s gonna asymptote towards that, right?

You’re never gonna know for sure,

but you can start to take out,

you can use good Bayesian approaches

and just keep updating and updating and updating

until you point to one sense of purpose.

So you wanna bound on how much novelty generation

it could be, got it.

So the ability to generate novelty

is correlated with high assembly index,

with assembly index?

Yeah, and yeah.

Cause the space of possibilities is bigger.

So that’s the key.

This could be a good, so I have a running joke

of like why Lex is single.

This could be a good part four.

So what you’re looking for in a robot partner

is ability to generate novelty.

And that’s, I suppose you would say

it’s a good definition of intelligence too.

Boy, is novelty a fuzzy concept.

Is creativity better?

Yeah, I mean, that’s all pretty fuzzy.

It’s kind of the same.

Maybe that’s why aliens haven’t come yet

is cause we’re not creating enough novelty.

Like there’s some kind of a hierarchy

of novelty in the universe.

Well, I think novelty is like things surprise you, right?

So it’s a very passive thing,

but I guess what I meant by saying creativity

is I think it’s much more active.

Like you think there’s like a mechanism

of like the things that exist are generating the creativity.

Novelty seems to be there’s some spontaneousness

and it’s completely decoupled from the things that exist.

No, I understand.

I think creativity is the mechanism

and novelty is the observable.


Novelty could just be surprising.

Your model of the world was broken

and not necessarily in a positive way.

That’s surprise.

So there’s three things now.

Let’s go back.

It’s cool.

All right, let’s go.

You got surprise, which is basically,

I mean, I’m surprised all the time

cause I don’t read very much.

I’m pretty dumb.

I was like, oh wow, this is,

I often used to invent new scientific, you know, ideas.

And I was really surprised by that.

And then I went and looked in literature properly

and it’s there.

So surprise, that’s to the extent

that you don’t have full information.

Creativity, the act of pushing on that kind of

on the causal structure and novelty,

which is measuring that degree, right?

So, and I think that’s pretty well defined in that regard.

So you want your robot, I mean,

and in the end, this is what I’m saying,

and in the end, that’s why actually the way the internet

and the printing press share some,

I actually think creativity has dropped a bit

since the internet because everyone’s just, you know,

just regurgitating stuff.

But of course, now it’s beginning to accelerate again

cause everyone’s using this tool to be creative

and boom, it’s exploding.

I think that’s what happens

when you create these new technologies.

That’s really helpful.

There’s a difference between novelty and surprise.

Okay, I think I was thinking about surprise.

If you give me a toy that surprises me for a bit,

it’d be great.

A robot that surprises me, you know.

An experiment that surprises you.

Yeah, I mean, that’s why I love doing experiments

cause I’m, I can’t.

It’s still exciting.

Yeah. Surprise is exciting.


Even negative surprise,

like some people love drama in relationships.

Like, it’s like, why the hell, why’d you do this?

That could be exciting too.

I could imagine companies selling updates

to their companion robots

that just basically generate negative surprise

just to spice things up a bit.

Yeah, it’s the push and pull.

That’s one of the components of love.

As you said, love is a complicated thing.

Oh, beauty.

I wanted to mention this cause you also tweeted,

I think this was Sarah.

No, it might’ve been Lee.

I don’t remember.

But it was a survey published in Nature

showing that scientists find.

That was me, yeah.


Anyway, there’s a plot.

This is published in Nature

of what scientists find beautiful in their work

and it separates biologists and physicists.

It’d be nice if you showed the full plot.

And there’s simplicity, elegance, hidden order,

inner logic of systems, symmetry, complexity, harmony,

and so on.

Is there any interesting things that stand out to you?

I think the fact that biologists like complexity

and pleasing colors.

Oh, there’s pleasing colors on there?

Yeah, yeah, yeah.

Or shapes.

Pleasing colors or shapes.

And then physicists obviously love simplicity

above all else.

Simplicity and elegance.

Simplicity, elegance.


They love symmetry.

And then biologists love complexity.

And, well, they just love a little bit less.

They love everything a little bit less,

but complexity a little bit more.

A little bit more.

That’s so interesting.

And pleasing colors or shapes.

Do you think it’s a useful,

I forget what your tweet was

that this is missing some of the something.

Oh, no, I think it’s because I think about

how explanations become causal to our future.

So I have this whole philosophy

that the theories we build

and the way we describe reality

should have the largest breadth of possibilities

for the future of what we can accomplish.

So in some sense, it’s not like Occam’s razor.

It’s not for simplicity.

It’s for optimism or the kind of future you can build.

And so I think you have to think this way

when you’re thinking about life and alien life,

because ultimately we’re trying to build,

I mean, science is just basically

our narratives about reality.

And now you’re building a narrative

that is what we are as physical systems.

It seems to me it needs to be as positive as possible

because it’s basically gonna shape

the future trajectory where we’re going.

And we don’t use that as a heuristic in theory building

because we think theories

are about predicting features of the world,

not causing them.

But if you look at the history of all of the development

of human thought, it’s caused the things that happen next.

So it’s not just about looking at the world

and observing it.

It’s about actually that feedback loop that’s missing.

And it’s not in any of those categories.

What do you think is the most beautiful idea

in the physics of life, in the chemistry of life,

in this, through all your exploration

with assembly theory, what is the thing

that made you step back and say this idea is beautiful

or potentially beautiful?

For me, it’s that the universe is a creative place.

I guess I want to think, and whether it’s true or not,

is that we are special in some way

and it’s not like an arbitrary added on epiphenomenon

or ad hoc feature of the universe that we exist,

but it’s something deep and intrinsic

to the structure of reality.

And to me, the most beautiful ideas that come out of that

is that the reason we exist is for the universe

to generate more things and to think about itself

and use that as a mechanism for creating more stuff.

That’s for me.

So like the life that this, however common it is,

is an intrinsic part, is a fundamental part

of this universe, at least, that we live in.

I think so.

I mean, it’s always interesting to me

because we have theories of quantum mechanics and gravity

and they’re supposed to be like

our most fundamental theories right now.

And they describe things like the interaction

of massive bodies or the way that charges accelerate

or all these kind of features.

And they’re these really deep theories

and they tell us a lot about how reality works,

but they’re completely agnostic to our existence.

And I can’t help but think that whatever describes us

has to be even deeper than that.

And I think incorporating memory, I guess,

causality, whatever the term you want to use

into the physics view of the world might be.

That’s the easiest way to do it.

It’s the cleanest, so here we go again

with the physicist, I’m a physicist.

The cleanest, I was gonna say the simplest,

most elegant way of resolving all of the kind of ways

that we have these paradoxes associated with life

when you, it’s not that life is not,

current physics is not incompatible with life,

but it doesn’t explain life.

And then you want to know where are the explanatory gaps

and this idea that we have an assembly

that time is fundamental and objects actually

are extended in time and have physical extent in time

is the cleanest way of resolving

a lot of the explanatory gaps.

So I’ve been, I struggled with assembly theory

for many years, because I could see this gap.

And I think when I first met Sarah

and we realized we were kind of talking

about the same problem, but we were,

we understood another language.

It was quite hilarious actually,

because it’s like, I have no idea what you’re talking about,

but I think it sounds right.

So for me, the most beautiful thing about assembly theory

is I realized the assembly theory explains

why the universe, why life is the universe

developing a memory, but not only that poetically,

I could actually go and measure it.

And I was like, holy shit, we would just,

we physically measured this thing,

this abstract thing, and we could measure it.

And not only could we measure it,

but we can then start to quantify the causal consequences.

Because, I mean, I think as a kind of inventing this

together with Sarah and her team, I thought

there was a quite a high chance that,

we’re doing science, there’s such a high probability

we’re wrong on this.

And I remember kind of trying to go to hard physicists,

mathematicians, complexity theorists,

and everyone just kind of giving me

such a hard time about it.

And so, this is kind of, you’ve just done this,

you’ve just done that, you’ve just recapitulated

an old theory.

And I was unable, I lacked the language to really explain,

and I had to, it was a real struggle.

So this realization that life, what life does

that physics cannot understand, or chemistry,

is the universe develops a memory

that’s causally actionable, and then we can measure it,

but it isn’t just a one thing,

there is this intrinsic property

of all the objects in the universe.

Like I’ve said before, but me holding up this water bottle,

it isn’t any other water bottle,

but it is a sum total of all the water bottles

that have existed, right?

And will likely change the future of water bottles

and for other objects.

So it’s this kind of, so for me,

assembly theory explains the soul in stuff.

The monology.

But it is the monology, it’s not like

Sheldrake’s morphic resonance,

where we have this kind of wooey thing

permeating the universe, it is the interaction

of objects of other objects.

And some objects have more instantaneous causal power,

that’s life, living things, and some objects

are the instantaneous output of that causal power,

dead objects, but they’re part of the lineage.

And that for me is fascinating and really beautiful.

And I think that even if we’re determined

to be totally wrong, I think it will help us,

help hopefully understand what life is

and go into tech life elsewhere and make life in the lab.

How does that make you feel, by the way?

Does it make you feel less special,

that you’re so deeply integrated,

interconnected to the lineage?

I mean, okay, on one level, I just wanted in my life

as a scientist, I wanted to have an interesting idea

just once or an original idea.

I mean, it was like, you know, so I think that was cool

that we had this idea and we were playing with it.

And I think also that I kind of, I mean,

it took me ages to realize that Sarah had also had

the same kind of form, coming towards the same formulation

just from a completely different point because I,

but no, it makes me feel special.

And it also makes me feel connected to the universe.

It also makes me feel not just humble about, you know,

being a living object in the universe,

but the fact that it makes me really optimistic

about what the universe is gonna do in the future

because we’re not just isolated phenomena, we are connected.

I will be able to have, you know, one of my small objectives

in life is to change the future of the universe

in some profound way, just by existing.

Yeah, that’s not ambitious at all.


I think it’s also good because it makes me feel less lonely

because I just realized I’m not like,

I mean, I’m a unique assembly structure,

but I have so much overlap with the other entities

I interact with that we’re not completely individual, right?

And yet your existence does have a huge amount of impact

on how this whole thing unrolls on the future of the world.

As individuals, that’s, yeah.

But I was gonna say. Local packets of agency.

I think we all have a profound impact on the future,

some more than others, right?

All human beings, all life.

And I mean, that’s why I think it’s a privilege

in a way for, you know, to say,

to assert some degree of ego and agency,

you know, I’m gonna make a computer

or make an origin life machine or we can do this thing.

But actually it’s just like, you know,

my life’s probably living, if there is a God

or there’s a soul in everything,

it’s really laughing at us going,

I fool these guys by giving them ego.

So they strive for this stuff and look what it does

for the assembly space of the universe.

And there’s always a possibility

that science can’t answer all of it.

So that part’s challenging for me.

There might be a limit to this thing.

Let me ask you a bunch of ridiculous questions

and I demand relatively short answers.

Lee, what’s the scariest thing you’ve ever done?

Or what’s a scary thing that pops to mind?

Giving seminars in front of other scientists.

That’s, yeah, that is terrifying.

I could, if I had more time,

I would ask you about the most embarrassing,

but we’ll spare you.

What about you Sarah, scariest thing?

Up there, some of the scary things you’ve done.

Actually the scariest for me was deciding

I wanted to get divorced

because it was like a totally radical like.

Life transformation.

Yeah, because we had been married for a really long time.

And I think it was just so much like,

I realized like so much of my individual agency

I didn’t realize I had before.

And that was just really like scary, like empowering scary,

but like terrifying.

Like you were living in a kind of one way

for your whole life.

And then you realized your life could be a different way.


Yeah, there’s a between humans.

I mean, that’s the beautiful thing about love

is the connection you have,

but it’s also becomes a dependency and breaking that.

Whether it’s a mentor, what’s your parents,

your close friends.

It’s almost like waking up.

Like just there’s a different reality.

Yeah, that was scary.

Reinventing yourself.

Okay, if you could, Lee,

maybe I’ll actually we’ll alternate.

Sarah, if you could be someone else for a day,

someone alive today,

you haven’t met yet, haven’t met yet.

Or maybe you could do one who you’ve met.

Who would it be?

Kim Kardashian.

No joke.

The woman’s brilliant.

I would just like to experience,

like I just, I think she’s got such an interesting

and very deep understanding of social reality.

But you also said you have appreciation

of love for fashion.

I do.

But that’s actually the same.

Like I just think it’s really interesting

because we live in a social reality

which is completely artificially constructed.

And some people are really genius about moving through that.

And I think she’s particularly good at it.

I wonder if she’s good at understanding or if she’s just.

I think it’s very deeply intrinsic to her.

So I don’t know if she has much.

She’s like surfing a wave.

How much cognitive awareness she has of it

or how strategic it is.

But I think it’s deeply fascinating.

So I guess that’s the first one that comes to mind.

What about you, Lee?

If you could be somebody for a day.

Don’t say Yoshua Bach.

Don’t say Kim Kardashian.

Let’s do it off the table.

Off the table.

No, I was gonna say I would like to be a,

does it have to be here today?

I was gonna say I’d like to be the latest arm processor.


I would like to be the latest arm processor.

I’d like to understand.

I would like to know what it feel like to basically.

You like being objects.

I like being objects.

I’ve always obsessed with being objects

ever since I was a kid.

What’s the best part of being an arm processor for a day?

I mean, I’d like to understand how I access my memory,

what it anticipates coming next in clock cycles.

What about how it feels like?

Yeah, I wanna know how it feels like to be.

To be useful.

Thanks for that.

All right, if, Lee, if everyone on Earth disappeared

and it was just you left, what would your days look like?

What would you do?

Nobody else left to impress.

Nobody, no, probably can’t really do any real science

at scale.

What would you do with your remaining days?

Get every possible tool I could

and put it in my workshop and just make stuff.

As, so try to make stuff.

Just try and make stuff.

Make companions.

I’m pretty much making companions probably, yeah.

So in the physical space.


What about you, Sarah?

What would you, when you’re just left alone on Earth,

you’re the last person.

Are there animals in this scenario?

No living beings.

No plants?

No plants.

Oh, interesting.

I was gonna say, I would just,

I would try to walk the entire planet,

at least all the landmass.

Well, that’s true.

So you probably don’t know if there’s stuff.

You could be searching for plants or other humans

or other animals. And what would I eat?

You just have daily just allotment.

I would just walk all the time, I think.

Of Soylent.

I don’t know why.

Just walk.

That’s just what came to mind.

You’re the explorer.

I would just walk.

And I guess I would make a goal of covering

all of the entire Earth.

Because what else are you gonna do with your time?

What’s an item on your bucket list, Sarah,

that you haven’t done yet, but you hope to do?


Travel to space.

I don’t know.

You know what’s funny with my bucket list?

I only know it was on my bucket list once I check it off.

Once you check it off.

So your bucket list is like a fog.

It’s like a mystery almost by doing it.

Yeah, so it’s very subconsciously driven.

So it’s in your subconscious in there.

I think so.

You’re bringing it to the surface.

I think most of the steering of our agency

is in our subconscious anyway,

so I just kind of go with the flow.

But I guess, no, seriously.

Yeah, no, I get it.

I don’t know.

I guess, but I would like to go on a submarine,

like to the bottom of the ocean.

I think that’d be really cool.

To the bottom of the ocean.

Are you captivated by the mystery of the ocean?

Like how little you know.

I am, yeah.

Yeah, what about you, Lee?

What item on your bucket list?

I don’t have a bucket list, but I’ve just made one.

I would love to take a computer to the moon or Mars

and make drugs off world.

Be the first chemist to make drugs off world.

The first drug manufacturer in space.

Yeah, why not?

Drugs in space.

Do they have to be somehow like be able to habitate,

like be able to survive on that particular space?

Or like what’s the connection between being on Mars

and doing many things?

I just would like to be there.

I would like to take the ability to have command

and control over chemicals programmatically off earth

to somewhere else in the universe.

That just seems like you like difficulty

engineering problems.

Before I die, if I can do that, that’s great.

Would you travel to space if you could?

Yeah, yeah, that’s what I’m saying.

I’d love to go into space, but not just to be a tourist.

I wanna take scientific experiment in space

and do a thing in space that’s never been done before.

That’s a real possibility.

Yeah, yeah, yeah.

So that’s why there’s no point in listening

things I can’t do, yeah.

All right, what small act of kindness

were you once shown that you will never forget?

Small act of kindness, not big.

Somebody was just kind to you.

Somebody did something sweet.

When I was a PhD student, someone helped me out

with just, basically, I needed a computer.

I needed some power, computation power,

and someone took pity on me and helped me.

I was really touched.

They didn’t have to.

And they were actually quite, they were a disabled scientist

and they had other things to do

rather than help some random PhD student.

Gave me access, taught me a lot of stuff.

Yeah, actually, when you’re a grad student

or when you’re a student, when you’re even a student,

the younger it is, the better.

The attention, the support, the love you get

from an older person, a teacher,

something like that is super powerful.

It’s fascinating.

And from the perspective of the teacher,

they might not realize the impact they have,

but that little bit, those few words,

a little bit of help can have a lot of impact.

What about you, Sarah?

Somebody give you a free Starbucks at some point?

I love free Starbucks.

I like it when you’re in the line at Starbucks

and somebody buys your coffee in front of you

and then you buy the next one.

I love those, but that’s not my example.

Those are great.

It makes me happy.

And now my kids get all excited when we do it,

when we go in, we’re the first ones in line doing it.

But I guess I can use a similar example

about just being a student.

So Paul Davies is a very well known theoretical physicist

and he was generous enough with his time

to take me on as a postdoc.

But before I became his postdoc,

he invited me to a workshop

at Arizona State University in the Beyond Center

and took a walk with me around campus

just to talk about ideas after.

And I think there were two things

that were completely generous about that.

One is Paul’s philosophy is always interacting

with young people.

It’s like you interact with a mind in the room.

It doesn’t matter how well known or whatever.

It’s like you evaluate the person for the person.

But he also gave me a book,

The Eerie Silence that he had written and he wrote in it,

This is How EE Gets to ET,

which was an antimeric excess,

which I worked on as a PhD student

was the origin of homochirality

all the way up to what the book was about,

which was are we alone in the universe

and is there intelligent life out there?

And it was just so much about the questions I wanted to ask

because it was just everything about,

just it was just really, really kind.

Like that it’s okay to ask these questions

and you can actually have strong enough to answer them.

A lot of my career is mostly his encouragement

to ask deep questions.

Like he gave me the space to do it

in ways that a lot of previous mentors had.

I mean, I’ve had a good experience with mentors,

but it was like go off the deep end,

ask the hardest questions.

And I think that’s the best gift you can give somebody.

What would you,

because you’re both fascinating minds

and not, I would say, nonstandard in the best possible way.

Is there advice you can give to young folks

how to be nonstandard, how to stand out,

novelty, how to generate novelty?

That’s what I want on my tombstone.

I have one.

He generated novelty.

No, no, how to.

Oh, how to.

How, how, still.

I just love doing science.

And so when I was younger, I was just,

just wanted to, I mean, I’m still not sure

I’m a real scientist, right?

So I want to try.

So my advice for the young people

is just, if you just, if you love asking questions,

then don’t be afraid to ask the question,

even if it pisses people off,

because if you piss people off,

you’re probably asking the right question.

What I would say though, is don’t do what I did,

which is just piss everyone off.

Try and work out how to, you know, I think,

if other people are challenged by your questions,

you will get not only respect,

but people will give you, create space for you,

because you’re doing something really new.

I really try to create space in my academic career,

with my team, really try and praise them

and push them to do new things.

So my advice is, try to do new things,

get feedback, and the universe will help you.


Because the universe likes novelty.

I think so.

I think so, right?

This one will keep them around.

Oh my God.

What about you, Sarah?

You too like to ask the really out there big questions.

Yeah, because I have a strong passion for them.

So I think it goes back to the love.

Like if you’re doing the thing you’re supposed to be doing,

you should really love it.

So I always tell people that they should do

the thing they’re most passionate about.

But I think a flip side of that is,

that’s when you become in some way,

like not to sound cheesy,

but like your best version of yourself.

So I guess for me, as I become more successful in my career,

I feel like I can be more myself as an individual.

And so there’s this, I’ve always been following

the questions I’m most interested in,

which very early on I was discouraged

from doing by many people

because they thought they were unanswerable questions.

And I always just thought,

well, if no one’s even trying to answer them,

of course they’re gonna be unanswerable.

And then that was kind of an odd viewpoint.

But the more I found my way in that space,

the more I also made a space for myself as a person,

because you’re basically generating the niche

that you wanna exist in.

And so I think that’s part of it,

is not just to follow your passion,

but also think about like,

who do you wanna be and create that?

Yeah, who am I, who do you wanna be?

I mean, yeah, play temporally with it.

Yeah, who am I now?

Who do I wanna be now?

But who do I wanna be in the future?

They’re not decoupled.

Yeah, I always wonder if that’s like,

if I become something, am I finding myself

or am I creating myself?

And I think those are somehow the same kind of thing.

I do feel often like I was always meant

to be this kind of thing.

But is that created or discovered?

I don’t know.

But basically go towards that direction.

If you were abducted by aliens, Sarah.

Excellent, I’m waiting, they can come find me.

They’re on a spaceship.

And then they somehow figured out the language you speak

and ask you, what are you?

What is, explain yourself.

Not you, Sarah, but the species.


Life on Earth, like we don’t have time.

We’re busy grad students from another planet.

I see.

What’s interesting about human civilization?

What’s interesting about you?

You specifically, too.

They could be very kind of personal, kind of pushy.

And yeah, how would you begin to describe?

Okay, I have one.

Because, you know, obviously I self identify

as a scientist and a physicist,

but intrinsically I feel more like an artist.

But it’s almost like you’re an artist

that you don’t know what you’re painting yet.

And I guess I feel like that’s humanity,

like in some sense, we’re creating something

I think is profound and potentially very beautiful

in existence of the universe.

But we’re just so, not night, we’re just early.

We’re early, we’re young.

We don’t know what we’re doing yet.

Yeah, what’s with the nuclear weapons?

That’d be a question, too.

Like what are you guys, why?

What are we doing with them?

This creativity that you talk, it sounds very nice,

but it’s, you’re seem to be.

We’re making things that are.

Like very destructive and like the rockets,

this seems very aggressive.

Yeah, I know.

This is my blinders on.

I don’t know.

I mean, it goes back to the whole conversation

about suffering.

I have a hard time regularizing certain aspects

of reality into what I wanna envision.

And that’s obviously problematic.

But nuclear power has also given us a lot of good things.


So both, that’s human nature.

Both human beings and the technology we create

has the capacity for evil and the capacity for good.

Yeah, and we can’t all be good all the time.

I mean, there’s like this huge misnomer

that you need to be liked by everyone universally.

And obviously that’s like an ideal,

but it’s physically impossible.

Like you can’t get a group of people in a room

and have everyone like each other all the time.

So I think that kind of tension is actually really important

that we have different aesthetics, different goals.

And sometimes conflict comes out of that.

Speaking of which, do Yuli and Yoshua Bach

ever say anything nice to each other?

Or is it always conflict?

We never have conflict.

We argue, but I don’t think arguments are bad.

It’s love.

I mean, I think the problem I have,

not problem, I think.

Here we go.

And he’s not here to defend himself.

No, I just, I don’t necessarily understand the,

I mean, he’s just talking at such a high level.

You know, I’m a dimwit, so I’m like, I spend some,

so I think a lot of our conflict is not conflict.

We actually, we actually have a, I think,

I mean, I can’t speak for Yash,

I have a deep appreciation for him, he’s brilliant.

But I think I’m kind of frustrated and I’m trying to,

he thinks the universe is a computer

and I want to turn the universe into a computer.

Yeah, that’s a small disagreement.

So what would you, how would you defend your life

to an alien when you’re being abducted?

Would you focus on the specifics of your life?

No, no, no, I would be, I would try to be

as random as possible and try and confuse them.

Oh, good, good.


That might be the wiser choice.

The Easter eggs in reality.

No, I mean, if aliens abducted me.

Would you play dumb?

No, no, no, I would try and be as random as possible.

I would try and do something that would surprise

the hell out of them, which I thought,

I mean, I’d probably risk it, they might kill me,

but I think that’d probably be funny.

That might, yeah, they might want to study you

for prolonged periods of time.

My reasoning is, if I wanted to stay alive, okay,

so if the thing is, if I wasn’t going back to Earth

and the job was to stay alive, if I could be

as surprising as possible, they’d keep me around

like a pet, right?

Petly, on the aliens page book.

So you’d be okay being a pet?

Well, no, but I mean.

The last human that survives would just be

a pet to the aliens.

I don’t know, but I mean, I think that might be fun

because then I might get some feedback

from their curiosity, but yeah.

Let me ask you this question.

Given our conversation has a very different meaning,

not a more profound meaning, perhaps,

but would you rather lose all of your old memories

or never be able to make new ones?

I would have to lose all my old memories.

Again, it’s the novelty.

What about you, Sarah?

I’m the same because I don’t think,

like, it’s about the future experience, right?

And in some sense, like you were saying earlier,

most of our lived experience is actually in our memories.

So if you can’t generate new memories,

it’s like you’re not alive anymore.

That’s it, yeah.

What comforts you on bad days?

When you look at human civilization,

when you look at your own life,

what gives you hope?

What makes you feel good about what we’re doing

about life at the small scale of you as a human

and at the big scale of us as a human civilization,

maybe the big scale of the universe?

Children, my kids.

But I also mean that in like a grand sense of like,

not a grand, but like future minds in some sense.

So for me, like the most bleak movie ever,

people worry about apocalyptic things

like AI existential risk and climate change,

which children of men.

The whole premise of the movie was

there can be no children born on the entire planet.

And the youngest person on the planet

is like 18 years old or something.

Like, can you imagine a world without children?

It’s just, it’s harrowing.

That’s the scariest thing.

So I think what gives me hope is always youth

and the hope of children

and the possibilities of the future they see.

And they grow up in a completely different reality

than adults do.

And I think we have a hard time seeing

what their reality actually looks like.

But I think most of the time it’s super interesting.

Yeah, they have dreams, they have imagination,

they have this kind of excitement.

It’s so cool, it’s so fun to watch.

And yeah, you feel like you’re almost getting in the way

of all that imagination.

What about you, Lee?

What gives you hope?

So when I go back to my eight year old self,

the thing that I dreamed of as my eight year old self

was this world in which technology became programmable

and there was the internet and I’d get information

and I would expand my consciousness by just,

you know, getting access to everything that was going on.

And this happened in my lifetime.

I mean, we really do have that.

I mean, okay, there’s some bad things.

You know, there’s TikTok, everyone just, whatever,

all the bad things about social media.

But I think,

I mean, I can’t quite believe my luck being born now.

So amazing.

Being able to program reality in some way.

Yeah, and the thing that I really find fascinating

about human beings is just how ingenious they are.

I’m, you know, whether it’s from my kids,

my research group, my peers, other companies,

just how ingenious everyone is.

And I’m pretty sure humanity has,

or our causal chain in which humanity is a vital part

in the future is gonna have a lot of fun.

And I’m just, yeah, it’s just mind blowing just to watch.

And, you know, so humans are ingenious

and I hope to help them be more ingenious if I can.

Well, what gives me hope,

what makes me feel good on bad days

is the existence of wild minds like yours,

novelty generators, assembly structures

that generate novelty and do so beautifully

and then tweet about it.

Sarah, I really, really enjoy talking to you.

I enjoy following you.

I’m a huge fan.

Sarah Lee, I hope to talk to you many times

in the future, maybe with Yoshua Bach.

You’re just incredible people.

Thank you for everything you do.

You’re awesome.

Thank you for talking today.

Really, really appreciate it.

Yeah, brilliant to be here.

Thanks for listening to this conversation

with Sarah Walker and Lee Cronin.

To support this podcast,

please check out our sponsors in the description.

Now, let me leave you with some words from Arthur C. Clarke.

Two possibilities exist.

Either we are alone in the universe or we are not.

Both are equally terrifying.

And let me, if I may, add to that

by saying that both possibilities, at least to me,

are both terrifying and exciting.

And keeping these two feelings in my heart

is a fun way to explore, to wonder, to think, and to live,

always a little bit on the edge of madness.

Thank you for listening.

I hope to see you next time.

comments powered by Disqus