back to indexDonald Hoffman: Reality is an Illusion - How Evolution Hid the Truth | Lex Fridman Podcast #293
link |
Whatever reality is, it's not what you see.
link |
What you see is just an adaptive fiction.
link |
The following is a conversation with Donald Hoffman,
link |
professor of cognitive sciences at UC Irvine,
link |
focusing his research on evolutionary psychology,
link |
visual perception, and consciousness.
link |
He's the author of over 120 scientific papers
link |
on these topics and his most recent book
link |
titled The Case Against Reality,
link |
Why Evolution Hid the Truth from Our Eyes.
link |
I think some of the most interesting ideas in this world,
link |
like those of Donald Hoffman's,
link |
attempt to shake the foundation
link |
of our understanding of reality,
link |
and thus they take a long time to internalize deeply.
link |
So proceed with caution.
link |
Questioning the fabric of reality
link |
can lead you to either madness or to truth.
link |
And the funny thing is, you won't know which is which.
link |
This is the Lex Friedman podcast.
link |
To support it, please check out our sponsors
link |
in the description.
link |
And now, dear friends, here's Donald Hoffman.
link |
In your book, The Case Against Reality,
link |
Why Evolution Hid the Truth from Our Eyes,
link |
you make the bold claim that the world we see
link |
with our eyes is not real.
link |
It's not even an abstraction of objective reality.
link |
It is completely detached from objective reality.
link |
Can you explain this idea?
link |
Right, so this is a theorem
link |
from evolution by natural selection.
link |
So the technical question that I and my team asked was,
link |
what is the probability that natural selection
link |
would shape sensory systems
link |
to see true properties of objective reality?
link |
And to our surprise,
link |
we found that the answer is precisely zero,
link |
except for one kind of structure
link |
that we can go into if you want to.
link |
But for any generic structure
link |
that you might think the world might have,
link |
a total order, a topology, metric,
link |
the probability is precisely zero
link |
that natural selection would shape any sensory system
link |
of any organism to see any aspect of objective reality.
link |
So in that sense, what we're seeing
link |
is what we need to see
link |
to stay alive long enough to reproduce.
link |
So in other words, we're seeing what we need
link |
to guide adaptive behavior, full stop.
link |
So the evolutionary process,
link |
the process that took us from the origin of life on Earth
link |
to the humans that we are today,
link |
that process does not maximize for truth,
link |
it maximizes for fitness, as you say, fitness beats truth.
link |
And fitness does not have to be connected to truth,
link |
And that's where you have an approach
link |
towards zero of probability
link |
that we have evolved human cognition,
link |
human consciousness, whatever it is,
link |
the magic that makes our mind work,
link |
evolved not for its ability to see the truth of reality,
link |
but its ability to survive in the environment.
link |
That's exactly right.
link |
So most of us intuitively think that surely
link |
the way that evolution will make our senses more fit
link |
is to make them tell us more truths,
link |
or at least the truths we need to know
link |
about objective reality, the truths we need in our niche.
link |
That's the standard view, and it was the view I took.
link |
I mean, that's sort of what we're taught
link |
or just even assume.
link |
It was just sort of like the intelligent assumption
link |
that we would all make.
link |
But we don't have to just wave our hands.
link |
Evolution of a natural selection
link |
is a mathematically precise theory.
link |
John Maynard Smith in the 70s
link |
created evolutionary game theory.
link |
And we have evolutionary graph theory
link |
and even genetic algorithms that we can use to study this.
link |
And so we don't have to wave our hands.
link |
It's a matter of theorem and proof and or simulation
link |
before you get the theorems and proofs.
link |
And a couple of graduate students of mine,
link |
Chester Mark and Brian Marion,
link |
did some wonderful simulations that tipped me off
link |
that there was something going on here.
link |
And then I went to a mathematician, Chetan Prakash,
link |
and Manish Singh, and some other friends of mine,
link |
But Chetan was the real mathematician behind all this.
link |
And he's proved several theorems
link |
that uniformly indicate that with one exception,
link |
which has to do with probability measures,
link |
there's no, the probability is zero.
link |
The reason there's an exception for probability measures,
link |
so called sigma algebras or sigma additive classes,
link |
is that for any scientific theory,
link |
there is the assumption that needs to be made
link |
that whatever structure,
link |
whatever probabilistic structure the world may have
link |
is not unrelated to the probabilistic structure
link |
of our perceptions.
link |
If they were completely unrelated,
link |
then no science would be possible.
link |
So this is technically the map from reality to our senses
link |
has to be a so called measurable map,
link |
has to preserve sigma algebras.
link |
But that means it could be infinite to one,
link |
and it could collapse all sorts of event information.
link |
But other than that, there's no requirement
link |
in standard evolutionary theory
link |
for fitness payoff functions, for example,
link |
to preserve any specific structures of objective reality.
link |
So you can ask the technical question.
link |
This is one of the avenues we took.
link |
If you look at all the fitness payoffs
link |
from whatever world structure you might want to imagine.
link |
So a world with say a total order on it.
link |
So it's got end states and they're totally ordered.
link |
And then you can have a set of maps from that world
link |
into a set of payoffs, say from zero to a thousand
link |
or whatever you want your payoffs to be.
link |
And you can just literally count all the payoff functions
link |
and just do the combinatorics and count them.
link |
And then you can ask the precise question,
link |
how many of those payoff functions preserve the total order?
link |
If that's what you're looking for,
link |
or how many preserve the topology?
link |
And you just count them and divide.
link |
So the number that are homomorphisms
link |
versus the total number, and then take the limit
link |
as the number of states in the world
link |
and the number of payoff values goes very large.
link |
And when you do that, you get zero every time.
link |
Okay, there's a million things to ask here.
link |
But first of all, just in case people
link |
are not familiar with your work,
link |
let's sort of linger on the big bold statement here,
link |
which is the thing we see with our eyes
link |
is not some kind of limited window into reality.
link |
It is completely detached from reality,
link |
likely completely detached from reality.
link |
You're saying 100% likely.
link |
Okay, so none of this is real in the way we think is real.
link |
In the way we have this intuition,
link |
there's like this table is some kind of abstraction,
link |
but underneath it all, there's atoms.
link |
And there's an entire century of physics
link |
that describes the functioning of those atoms
link |
and the quirks that make them up.
link |
There's many Nobel Prizes about particles and fields
link |
and all that kind of stuff that slowly builds up
link |
to something that's perceivable to us,
link |
both with our eyes, with our different senses as this table.
link |
Then there's also ideas of chemistry
link |
that over layers of abstraction, from DNA to embryos,
link |
the cells that make the human body.
link |
So all of that is not real.
link |
It's a real experience,
link |
and it's a real adaptive set of perceptions.
link |
So it's an adaptive set of perceptions, full stop.
link |
We want to think that the perceptions are real.
link |
So their perceptions are real as perceptions, right?
link |
We are having our perceptions,
link |
but we've assumed that there's a pretty tight relationship
link |
between our perceptions and reality.
link |
If I look up and see the moon,
link |
then there is something that exists in space and time
link |
that matches what I perceive.
link |
And all I'm saying is that if you take evolution
link |
by natural selection seriously, then that is precluded.
link |
That our perceptions are there.
link |
They're there to guide adaptive behavior, full stop.
link |
They're not there to show you the truth.
link |
In fact, the way I think about it is
link |
they're there to hide the truth
link |
because the truth is too complicated.
link |
It's just like if you're trying to use your laptop
link |
to write an email, right?
link |
What you're doing is toggling voltages in the computer,
link |
but good luck trying to do it that way.
link |
The reason why we have a user interface
link |
is because we don't want to know that quote unquote truth,
link |
the diodes and resistors and all that terrible hardware.
link |
If you had to know all that truth,
link |
your friends wouldn't hear from you.
link |
So what evolution gave us was perceptions
link |
that guide adaptive behavior.
link |
And part of that process, it turns out,
link |
means hiding the truth and giving you eye candy.
link |
So what's the difference between hiding the truth
link |
and forming abstractions,
link |
layers upon layers of abstractions
link |
over low level voltages and transistors
link |
and chips and programming languages
link |
from assembly to Python that then leads you
link |
to be able to have an interface like Chrome
link |
where you open up another set of JavaScript and HTML
link |
programming languages that lead you
link |
to have a graphical user interface
link |
and which you can then send your friends an email.
link |
Is that completely detached from the zeros and ones
link |
that are firing away inside the computer?
link |
Of course, when I talk about the user interface
link |
on your desktop, there's this whole sophisticated
link |
backstory to it, right?
link |
That the hardware and the software
link |
that's allowing that to happen.
link |
Evolution doesn't tell us the backstory, right?
link |
So the theory of evolution is not going to be adequate
link |
to tell you what is that backstory.
link |
It's gonna say that whatever reality is,
link |
and that's the interesting thing,
link |
it says whatever reality is, you don't see it.
link |
You see a user interface,
link |
but it doesn't tell you what that user interface is,
link |
how it's built, right?
link |
Now, we can try to look at certain aspects
link |
of the interface, but already we're gonna look at that
link |
and go, okay, before I would look at neurons
link |
and I was assuming that I was seeing something
link |
that was at least partially true.
link |
And now I'm realizing that it could be like looking
link |
at the pixels on my desktop or icons on my desktop
link |
and good luck going from that to the data structures
link |
and then the voltages and I mean, good luck.
link |
There's just no way.
link |
So what's interesting about this is that
link |
our scientific theories are precise enough
link |
and rigorous enough to tell us certain limits,
link |
but, and even limits of the theories themselves,
link |
but they're not going to tell us what the next move is
link |
and that's where scientific creativity comes in.
link |
So the stuff that I'm saying here, for example,
link |
is not alien to physicists.
link |
The physicists are saying precisely the same thing
link |
that space time is doomed.
link |
We've assumed that space time is fundamental.
link |
We've assumed that for several centuries
link |
and it's been very useful.
link |
So all the things that you were mentioning,
link |
the particles and all the work that's been done,
link |
that's all been done in space time,
link |
but now physicists are saying space time is doomed.
link |
There's no such thing as space time fundamentally
link |
in the laws of physics.
link |
And that comes actually out of gravity
link |
together with quantum field theory,
link |
which just comes right out of it.
link |
It's a theorem of those two theories put together,
link |
but it doesn't tell you what's behind it.
link |
So the physicists know that their best theories,
link |
Einstein's gravity and quantum field theory put together
link |
entail that space time cannot be fundamental
link |
and therefore particles in space time cannot be fundamental.
link |
They're just irreducible representations
link |
of the symmetries of space time.
link |
That's what they are.
link |
So we have, so space time, so we put the two together.
link |
We put together what the physicists are discovering
link |
and we can talk about how they do that.
link |
And then we, the new discoveries
link |
from evolution of a natural selection.
link |
Both of these discoveries are really in the last 20 years.
link |
And what both are saying is space time
link |
has had a good ride.
link |
It's been very useful.
link |
Reductionism has been useful, but it's over.
link |
And it's time for us to go beyond.
link |
When you say space time is doomed,
link |
is it the space, is it the time,
link |
is it the very hard coded specification of four dimensions?
link |
Or are you specifically referring
link |
to the kind of perceptual domain
link |
that humans operate in, which is space time?
link |
You think like there's a 3D, like our world
link |
is three dimensional and time progresses forward.
link |
Therefore, three dimensions plus one, 4D.
link |
What exactly do you mean by space time?
link |
And what do you mean by space time is doomed?
link |
So this is, by the way, not my quote.
link |
This is from, for example, Nima Arkanihaim Ed
link |
at the Institute for Advanced Study at Princeton.
link |
Ed Witten, also there.
link |
David Gross, Nobel Prize winner.
link |
So this is not just something the cognitive scientists,
link |
this is what the physicists are saying.
link |
Yeah, the physicists, they're space time skeptics.
link |
Well, yeah, they're saying that,
link |
and I can say exactly why they think it's doomed.
link |
But what they're saying is that,
link |
because your question was what aspect of space time,
link |
what are we talking about here?
link |
It's both space and time.
link |
They're union into space time as an Einstein's theory.
link |
And they're basically saying that even quantum theory,
link |
this is with Nima Arkanihaim Ed, especially.
link |
So Hilbert spaces will not be fundamental either.
link |
So that the notion of Hilbert space,
link |
which is really critical to quantum field theory,
link |
quantum information theory,
link |
that's not going to figure
link |
in the fundamental new laws of physics.
link |
So what they're looking for
link |
is some new mathematical structures beyond space time,
link |
beyond Einstein's four dimensional space time
link |
or super symmetric version,
link |
geometric algebra signature two comma four kind of.
link |
There are different ways that you can represent it,
link |
but they're finding new structures.
link |
And then by the way, they're succeeding now.
link |
They're finding, they found something
link |
called the amplituhedron.
link |
This is Nima and his colleagues,
link |
the cosmological polytope.
link |
So there are these like polytopes,
link |
these polyhedra in multi dimensions,
link |
generalizations of simplices that are coding for,
link |
for example, the scattering amplitudes of processes
link |
in the Large Hadron Collider and other colliders.
link |
So they're finding that if they let go of space time,
link |
completely, they're finding new ways
link |
of computing these scattering amplitudes
link |
that turn literally billions of terms into one term.
link |
When you do it in space and time,
link |
because it's the wrong framework,
link |
it's just a user interface from,
link |
that's not from the evolutionary point of view,
link |
it's just user interface.
link |
It's not a deep insight into the nature of reality.
link |
So it's missing deep symmetry
link |
is something called a dual conformal symmetry,
link |
which turns out to be true of the scattering data,
link |
but you can't see it in space time.
link |
And it's making the computations way too complicated
link |
because you're trying to compute all the loops
link |
in the Feynman diagrams and all the Feynman integrals.
link |
So see the Feynman approach to the scattering amplitudes
link |
is trying to enforce two critical properties of space time,
link |
locality and unitarity.
link |
And so by, when you enforce those,
link |
you get all these loops and multiple,
link |
different levels of loops.
link |
And for each of those,
link |
you have to add new terms to your computation.
link |
But when you do it outside of space time,
link |
you don't have the notion of unitarity.
link |
You don't have the notion of locality.
link |
You have something deeper
link |
and it's capturing some symmetries
link |
that are actually true of the data.
link |
And, but then when you look at the geometry
link |
of the facets of these polytopes,
link |
then certain of them will code for unitarity and locality.
link |
So it actually comes out of the structure
link |
of these deep polytopes.
link |
So what we're finding is there's this whole new world.
link |
Now beyond space time that is making explicit symmetries
link |
that are true of the data
link |
that cannot be seen in space time.
link |
And that is turning the computations
link |
from billions of terms to one or two or a handful of terms.
link |
So we're getting insights into symmetries
link |
and all of a sudden the math is becoming simple
link |
because we're not doing something silly.
link |
We're not adding up all these loops in space time.
link |
We're doing something far deeper.
link |
But they don't know what this world is about.
link |
Also, they're in an interesting position
link |
where we know that space time is doomed.
link |
And I should probably tell you why it's doomed,
link |
what they're saying about why it's doomed.
link |
But they need a flashlight to look beyond space time.
link |
What flashlight are we gonna use
link |
to look into the dark beyond space time?
link |
Because Einstein's theory and quantum theory
link |
can't tell us what's beyond them.
link |
All they can do is tell us that when you put us together,
link |
space time is doomed at 10 to the minus 33 centimeters,
link |
10 to the minus 43 seconds.
link |
Beyond that, space time doesn't even make sense.
link |
It just has no operational definition.
link |
So, but it doesn't tell you what's beyond.
link |
And so they're just looking for deep structures
link |
like guessing is really fun.
link |
So these really brilliant guys, generic brilliant men
link |
and women who are doing this work, physicists,
link |
are making guesses about these structures,
link |
informed guesses, because they're trying to ask,
link |
well, okay, what deeper structure could give us
link |
the stuff that we're seeing in space time,
link |
but without certain commitments
link |
that we have to make in space time, like locality.
link |
So they make these brilliant guesses.
link |
And of course, most of the time you're gonna be wrong,
link |
but once you get one or two that start to pay off
link |
and then you get some lucky breaks.
link |
So they got a lucky break back in 1986.
link |
Couple of mathematicians named Park and Taylor
link |
took the scattering amplitude for two gluons coming in
link |
at high energy and four gluons going out at low energy.
link |
So that kind of scattering thing.
link |
So it's like apparently for people who are into this,
link |
that's sort of something that happens so often
link |
you need to be able to find it and get rid of those
link |
cause you already know about that and you need to.
link |
So you needed to compute them.
link |
It was billions of terms and they couldn't do it
link |
even though for the supercomputers couldn't do that
link |
for the many billions or millions of times per second
link |
they needed to do it.
link |
So the experimentals begged the theorists,
link |
please, you got it.
link |
And so Park and Taylor took the billions of terms,
link |
hundreds of pages and miraculously turned it into nine.
link |
And then a little bit later,
link |
they guessed one term expression
link |
that turned out to be equivalent.
link |
So billions of terms reduced to one term,
link |
that so called famous Park Taylor formula, 1986.
link |
And that was like, okay, where did that come from?
link |
This is a pointer into a deep realm, beyond space and time,
link |
but no one, I mean, what can you do with it?
link |
And they thought maybe it was a one off,
link |
but then other formulas started coming up.
link |
And then eventually Neymar, Connie, Hamid and his team
link |
found this thing called the amplituhedron,
link |
which really sort of captures the whole,
link |
a big part of the whole ball of wax.
link |
I'm sure they would say, no, there's plenty more to do.
link |
So I won't say they did it all by any means.
link |
They're looking at the cosmological polytope as well.
link |
So what's remarkable to me is that two pillars
link |
of modern science, quantum field theory with gravity
link |
on the one hand and evolution by natural selection
link |
on the other, just in the last 20 years
link |
have very clearly said space time has had a good run.
link |
Reductionism has been a fantastic methodology.
link |
So we had a great ontology of space time,
link |
a great methodology of reductionism.
link |
Now it's time for a new trick.
link |
But now you need to go deeper and show,
link |
but by the way, this doesn't mean we throw away
link |
everything we've done, not by a long shot.
link |
Every new idea that we come up with beyond space time
link |
must project precisely into space time.
link |
And it better give us back everything that we know
link |
and love in space time or generalizations,
link |
or it's not gonna be taken seriously and it shouldn't be.
link |
So we have a strong constraint on whatever we're going to do
link |
beyond space time, it needs to project into space time.
link |
And whatever this deeper theory is,
link |
it may not itself have evolution by natural selection.
link |
This may not be part of this deeper realm.
link |
But when we take whatever that thing is beyond space time
link |
and project it into space time,
link |
it has to look like evolution by natural selection
link |
So that's a strong constraint on this work.
link |
So even the evolution by natural selection
link |
and quantum field theory could be interfaces
link |
into something that doesn't look anything like,
link |
like you mentioned.
link |
I mean, it's interesting to think that evolution
link |
might be a very crappy interface
link |
into something much deeper.
link |
They're both telling us that the framework that you've had
link |
can only go so far and it has to stop.
link |
And there's something beyond.
link |
And the very framework that is space and time itself.
link |
Now, of course, evolution by natural selection
link |
is not telling us about like Einstein's relativistic
link |
So that was another question you asked a little bit earlier.
link |
It's telling us more about our perceptual space and time,
link |
which we have used as the basis for creating
link |
first Newtonian space versus time
link |
as a mathematical extension of our perceptions.
link |
And then Einstein then took that and extended it even further.
link |
So the relationship between what evolution is telling us
link |
and what the physicists are telling us is that
link |
in some sense, the Newton and Einstein space time
link |
are formulated as sort of rigorous extensions
link |
of our perceptual space,
link |
making it mathematically rigorous
link |
and laying out the symmetries that they find there.
link |
So that's sort of the relationship between them.
link |
So it's the perceptual space time
link |
that evolution is telling us
link |
is just a user interface effectively.
link |
And then the physicists are finding
link |
that even the mathematical extension of that
link |
into the Einsteinian formulation has to be as well,
link |
not the final story, there's something deeper.
link |
So let me ask you about reductionism and interfaces
link |
as we march forward from Newtonian physics
link |
to quantum mechanics.
link |
These are all, in your view, interfaces.
link |
Are we getting closer to objective reality?
link |
How do we know if these interfaces in the process of science,
link |
the reason we like those interfaces
link |
is because they're predictive of some aspects,
link |
strongly predictive about some aspects of our reality.
link |
Is that completely deviating
link |
from our understanding of that reality
link |
or is it helping us get closer and closer and closer?
link |
Well, of course, one critical constraint
link |
on all of our theories
link |
is that they are empirically tested
link |
and pass the experiments that we have for them.
link |
So no one's arguing against experiments being important
link |
and wanting to test all of our current theories
link |
and any new theories on that.
link |
So that's all there.
link |
But we have good reason to believe
link |
that science will never get a theory of everything.
link |
Everything, everything.
link |
Everything, everything, right.
link |
A final theory of everything, right.
link |
I think that my own take is, for what it's worth,
link |
is that Gödel's incompleteness theorem
link |
sort of points us in that direction,
link |
that even with mathematics,
link |
any finite axiomatization that's sophisticated enough
link |
to be able to do arithmetic,
link |
it's easy to show that there'll be statements that are true,
link |
that can't be proven,
link |
can't be deduced from within that framework.
link |
And if you add the new statements to your axioms,
link |
then there'll be always new statements that are true,
link |
but can't be proven with a new axiom system.
link |
And the best scientific theories in physics, for example,
link |
and also now evolution, are mathematical.
link |
So our theories are gonna be,
link |
they're gonna have their own assumptions
link |
and they'll be mathematically precise.
link |
And there'll be theories, perhaps,
link |
of everything except those assumptions,
link |
because the assumptions are,
link |
we say, please grant me these assumptions.
link |
If you grant me these assumptions,
link |
then I can explain this other stuff.
link |
So you have the assumptions that are like miracles,
link |
as far as the theory is concerned.
link |
They're not explained.
link |
They're the starting points for explanation.
link |
And then you have the mathematical structure
link |
of the theory itself, which will have the Gödel limits.
link |
And so my take is that reality,
link |
reality, whatever it is, is always going to transcend
link |
any conceptual theory that we didn't come up with.
link |
There's always gonna be mystery at the edges.
link |
Contradictions and all that kind of stuff.
link |
So there's this idea that is brought up
link |
in the financial space of settlement of transactions.
link |
It's often talked about in cryptocurrency, especially.
link |
So you could do, you know, money, cash,
link |
is not connected to anything.
link |
It used to be connected to gold, to physical reality,
link |
but then you can use money to exchange,
link |
to exchange value, to transact.
link |
So when it was on the gold standard,
link |
the money would represent some stable component of reality.
link |
Isn't it more effective to avoid things like hyperinflation
link |
if we generalize that idea?
link |
Isn't it better to connect your,
link |
whatever we humans are doing
link |
in the social interaction space with each other,
link |
isn't it better from an evolutionary perspective
link |
to connect it to some degree to reality
link |
so that the transactions are settled
link |
with something that's universal,
link |
as opposed to us constantly operating
link |
in something that's a complete illusion?
link |
Isn't it easy to hyperinflate that?
link |
Like where you really deviate very, very far away
link |
from the underlying reality,
link |
or do you not never get in trouble for this?
link |
Can you just completely drift far, far away
link |
from the underlying reality and never get in trouble?
link |
That's a great question, on the financial side,
link |
there's two levels at least
link |
that we could take your question.
link |
One is strictly like evolutionary psychology
link |
of financial systems, and that's pretty interesting.
link |
And there the decentralized idea,
link |
the DeFi kind of idea in cryptocurrencies
link |
may make good sense
link |
from just an evolutionary psychology point of view.
link |
Having human nature being what it is,
link |
putting a lot of faith in a few central controllers
link |
depends a lot on the veracity of those
link |
and trustworthiness of those few central controllers.
link |
And we have ample evidence time and again
link |
that that's often betrayed.
link |
So it makes good evolutionary sense, I would say,
link |
to have a decentralized,
link |
I mean, democracy is a step in that direction, right?
link |
We don't have a monarch now telling us what to do,
link |
we decentralize things, right?
link |
Because if the monarch,
link |
if you have Marcus Aurelius as your emperor, you're great.
link |
If you have Nero, it's not so great.
link |
And so we don't want that.
link |
So democracy is a step in that direction,
link |
but I think the DeFi thing is an even bigger step
link |
and is going to even make the democratization even greater.
link |
So that's one level of it.
link |
Also, the fact that power corrupts
link |
and absolute power corrupts absolutely
link |
is also a consequence of evolution.
link |
That's also a feature, I think, right?
link |
You can argue from the long span of living organisms,
link |
it's nice for power to corrupt for you to,
link |
so mad men and women throughout history
link |
might be useful to teach us a lesson about ourselves.
link |
We can learn from our negative example, right?
link |
Right, right, right.
link |
Power does corrupt and I think that you can think about that
link |
again from an evolutionary point of view.
link |
But I think that your question was a little deeper
link |
when that was, does the evolutionary interface idea
link |
sort of unhinge science from some kind of important test
link |
for the theories, right?
link |
We don't want, it doesn't mean that anything goes
link |
in scientific theory, but there's no,
link |
if we don't see the truth,
link |
is there no way to tether our theories and test them?
link |
And I think there's no problem there.
link |
We can only test things in terms of what we can measure
link |
with our senses in space and time.
link |
So we're going to have to continue to do experiments
link |
and, but we're going to re,
link |
we're going to understand a little bit differently
link |
what those experiments are.
link |
We had thought that when we see a pointer
link |
on some machine in an experiment,
link |
that the machine exists, the pointer exists
link |
and the values exist even when no one is looking at them
link |
and that they're an objective truth.
link |
And our best theories are telling us no,
link |
the pointers are just pointers
link |
and that's what you have to rely on
link |
for making your judgments.
link |
But even the pointers themselves
link |
are not the objective reality.
link |
So, and I think Gödel is telling us that,
link |
not that anything goes, but as you develop
link |
new axiom systems, you will find out what goes
link |
within that axiom system
link |
and what testable predictions you can make.
link |
So I don't think we're untethered.
link |
We continue to do experiments.
link |
What I think we won't have that we want
link |
is a conceptual understanding
link |
that gives us a theory of everything
link |
that's final and complete.
link |
I think that this is, to put it another way,
link |
this is job security for scientists.
link |
Our job will never be done.
link |
It's job security for neuroscience.
link |
Because before we thought that when we looked in the brain,
link |
we saw neurons and neural networks
link |
and action potentials and synapses and so forth.
link |
And that was it, that was the reality.
link |
Now we have to reverse engineer that.
link |
We have to say, what is beyond space time?
link |
What is a dynamical system beyond space time?
link |
That when we project it into Einstein's space time,
link |
gives us things that look like neurons
link |
and neural networks and synapses.
link |
So we have to reverse engineer it.
link |
So there's gonna be lots more work for neuroscience.
link |
It's gonna be far more complicated
link |
and difficult and challenging.
link |
But that's wonderful, that's what we need to do.
link |
We thought neurons exist when they are perceived
link |
In the same way that if I show you,
link |
when I say they don't exist,
link |
I should be very, very concrete.
link |
If I draw on a piece of paper,
link |
a little sketch of something that is called the Necker cube,
link |
it's just a little line drawing of a cube, right?
link |
It's not a flat piece of paper.
link |
If I execute it well, and I show it to you,
link |
you'll see a 3D cube and you'll see it flip.
link |
Sometimes you'll see one face in front,
link |
sometimes you'll see the other face in front.
link |
But if I ask you, which face is in front
link |
when you don't look?
link |
The answer is, well, neither face is in front
link |
because there's no cube.
link |
There's just a flat piece of paper.
link |
So when you look at the piece of paper,
link |
you perceptually create the cube.
link |
And when you look at it,
link |
then you fix one face to be in front and one face to be.
link |
So that's what I mean when I say it doesn't exist.
link |
Space time itself is like the cube.
link |
It's a data structure that your sensory systems construct,
link |
whatever your sensory systems mean now,
link |
because we now have to not even take that for granted.
link |
But there are perceptions that you construct on the fly
link |
and they're data structures in a computer science sense,
link |
and you garbage collect them when you don't need them.
link |
So you create them and garbage collect them.
link |
But is it possible that it's mapped well
link |
in some concrete, predictable way to objective reality?
link |
The sheet of paper, this two dimensional space,
link |
or we can talk about space time,
link |
maps in some way that we maybe don't yet understand,
link |
but we'll one day understand what that mapping is,
link |
but it maps reliably.
link |
It is tethered in that way.
link |
And so the new theories that the physicists are finding
link |
beyond space time have that kind of tethering.
link |
So they show precisely how you start with an epileptic hedron
link |
and how you project this high dimensional structure
link |
into the four dimensions of space time.
link |
So there's a precise procedure that relates the two.
link |
And they're doing the same thing
link |
with the cosmological polytopes.
link |
So they're the ones that are making the most concrete
link |
and fun advances going beyond space time.
link |
And they're tethering it, right?
link |
They say this is precisely the mathematical projection
link |
from this deeper structure into space time.
link |
One thing I'll say about, as a non physicist,
link |
what I find interesting is that they're finding just geometry,
link |
but there's no notion of dynamics.
link |
Right now, they're just finding
link |
these static geometric structures, which is impressive.
link |
So I'm not putting them down.
link |
This is what they're doing is unbelievably complicated
link |
and brilliant and adventurous, it's all those things.
link |
And beautiful from a human aesthetic perspective
link |
because geometry is beautiful.
link |
And they're finding symmetries that are true of the data
link |
that can't be seen in space time.
link |
But I'm looking for a theory beyond space time
link |
that's a dynamical theory.
link |
I would love to find, and we can talk about that
link |
at some point, a theory of consciousness
link |
in which the dynamics of consciousness itself
link |
will give rise to the geometry
link |
that the physicists are finding beyond space time.
link |
If we can do that,
link |
then we'd have a completely different way
link |
of looking at how consciousness is related
link |
to what we call the brain or the physical world
link |
more generally, right?
link |
Right now, all of my brilliant colleagues,
link |
well, 99% of them are trying to,
link |
they're assuming space time is fundamental.
link |
They're assuming that particles are fundamental,
link |
quarks, gluons, leptons, and so forth.
link |
Elements, atoms, and so forth are fundamental
link |
and that therefore neurons and brains
link |
are part of objective reality.
link |
And that somehow when you get matter
link |
that's complicated enough,
link |
it will somehow generate conscious experiences
link |
by its functional properties.
link |
Or if you're panpsychist, maybe you,
link |
in addition to the physical properties of particles,
link |
you add your consciousness property as well.
link |
And then you combine these physical and conscious properties
link |
to get more complicated ones.
link |
But they're all doing it within space time.
link |
All of the work that's being done on consciousness
link |
and its relationship to the brain
link |
is all assumed something that our best theories
link |
are telling us is doomed, space time.
link |
Why does that particular assumption bother you the most?
link |
So you bring up space time.
link |
I mean, that's just one useful interface
link |
we've used for a long time.
link |
Surely there's other interfaces.
link |
Is space time just one of the big ones
link |
that you, to build up people's intuition
link |
about the fact that they do assume a lot of things strongly?
link |
Or is it in fact the fundamental flaw
link |
in the way we see the world?
link |
Well, everything else that we think we know
link |
are things in space time.
link |
And so when you say space time is doomed,
link |
this is a shot to the heart of the whole framework,
link |
the whole conceptual framework that we've had in science.
link |
Not to the scientific method,
link |
but to the fundamental ontology
link |
and also the fundamental methodology,
link |
the ontology of space time and its contents,
link |
and the methodology of reductionism,
link |
which is that as we go to smaller scales in space time,
link |
we will find more and more fundamental laws.
link |
And that's been very useful for space and time for centuries,
link |
reductionism for centuries.
link |
But now we realize that that's over.
link |
Reductionism is in fact dead, as is space time.
link |
What exactly is reductionism?
link |
What is the process of reductionism
link |
that is different than some of the physicists
link |
that you mentioned that are trying to think,
link |
trying to let go of the assumption of space time?
link |
Looking beyond, isn't that still trying to come up
link |
with a simple model that explains this whole thing?
link |
Isn't it still reducing?
link |
It's a wonderful question,
link |
because it really helps to clarify two different notions,
link |
which is scientific explanation on the one hand,
link |
and a particular kind of scientific explanation on the other,
link |
which is the reductionist.
link |
So the reductionist explanation is saying,
link |
I will start with things that are smaller in space time
link |
and therefore more fundamental,
link |
where the laws are more fundamental.
link |
So we go to just smaller and smaller scales.
link |
Whereas in science more generally,
link |
we just say like when Einstein
link |
did the special theory of relativity,
link |
he's saying, let me have a couple of postulates.
link |
I will assume that the speed of light is universal
link |
for all observers in uniform motion,
link |
and that the laws of physics,
link |
so if you're for uniform motion are,
link |
that's not a reductionist.
link |
Those are saying, grant me these assumptions.
link |
I can build this entire concept of space time out of it.
link |
It's not a reductionist thing.
link |
You're not going to smaller and smaller scales of space.
link |
You're coming up with these deep, deep principles.
link |
Same thing with his theory of gravity, right?
link |
It's the falling elevator idea, right?
link |
So this is not a reductionist kind of thing.
link |
It's something different.
link |
So simplification is a bigger thing than just reductionism.
link |
Reductionism has been a particularly useful
link |
kind of scientific explanation,
link |
for example, in thermodynamics, right?
link |
Where the notion that we have of heat,
link |
some macroscopic thing like temperature and heat,
link |
it turns out that Neil Boltzmann and others discovered,
link |
well, hey, if we go to smaller and smaller scales,
link |
we find these things called molecules or atoms.
link |
And if we think of them as bouncing around
link |
and having some kind of energy,
link |
then what we call heat really can be reduced to that.
link |
And so that's a particularly useful kind of reduction,
link |
is a useful kind of scientific explanation
link |
that works within a range of scales within space time.
link |
But we know now precisely where that has to stop.
link |
At 10 to the minus 33 centimeters
link |
and 10 to the minus 43 seconds.
link |
And I would be impressed
link |
if it was 10 to the minus 33 trillion centimeters.
link |
I'm not terribly impressed at 10 to the minus 33 centimeters.
link |
I don't even know how to comprehend
link |
either of those numbers, frankly.
link |
Just a small aside,
link |
because I am a computer science person,
link |
I also find cellular automata beautiful.
link |
And so you have somebody like Stephen Wolfram,
link |
who recently has been very excitedly exploring
link |
a proposal for a data structure
link |
that could be the numbers that would make you
link |
a little bit happier in terms of scale,
link |
because they're very, very, very, very tiny.
link |
So do you like this space of exploration
link |
of really thinking, letting go of space time,
link |
letting go of everything and trying to think
link |
what kind of data structures
link |
could be underneath this whole mess?
link |
So if they're thinking about these as outside of space time,
link |
then that's what we have to do.
link |
That's what our best theories are telling us.
link |
You now have to think outside of space time.
link |
Now, of course, I should back up and say,
link |
we know that Einstein surpassed Newton, right?
link |
But that doesn't mean that there's not good work
link |
There's all sorts of Newtonian physics
link |
that takes us to the moon and so forth,
link |
and there's lots of good problems
link |
that we want to solve with Newtonian physics.
link |
The same thing will be true of space time.
link |
It's not like we're gonna stop using space time.
link |
We'll continue to do all sorts of good work there.
link |
But for those scientists who are really looking
link |
to go deeper, to actually find the next,
link |
just like what Einstein did to Newton,
link |
what are we gonna do to Einstein?
link |
How do we get beyond Einstein and quantum theory
link |
to something deeper?
link |
Then we have to actually let go.
link |
And if we're gonna do like this automata kind of approach,
link |
it's critical that it's not automata in space time,
link |
it's automata prior to space time,
link |
from which we're going to show how space time emerges.
link |
If you're doing automata within space time,
link |
well, that might be a fun model,
link |
but it's not the radical new step that we need.
link |
Yeah, so the space time emerges from that whatever system.
link |
Like you're saying, it's a dynamical system.
link |
Do we even have an understanding what dynamical means
link |
when we go beyond?
link |
When you start to think about dynamics,
link |
it could mean a lot of things.
link |
Even causality could mean a lot of things
link |
if we realize that everything's an interface.
link |
Like how much do we really know is an interesting question.
link |
Because you brought up neurons,
link |
I gotta ask you yet another tangent.
link |
There's a paper I remember a while ago looking at
link |
called Could a Neuroscientist Understand a Microprocessor?
link |
And I just enjoyed that thought experiment
link |
that they provided, which is they basically,
link |
it's a couple of neuroscientists,
link |
Eric Jonas and Conrad Cording,
link |
who use the tools of neuroscience
link |
to analyze a microprocessor, so a computer chip.
link |
Now, if we lesion it here, what happens and so forth,
link |
and if you go and lesion in a computer,
link |
it's very, very clear that lesion experiments on computers
link |
are not gonna give you a lot of insight into how it works.
link |
And also the measurement devices and the kind of sort of,
link |
just using the basic approaches of neuroscience,
link |
collecting the data, trying to intuit
link |
about the underlying function of it.
link |
And that helps you understand that
link |
our scientific exploration of concepts,
link |
depending on the field,
link |
are maybe in the very, very early stages.
link |
I wouldn't say it leaves us astray,
link |
perhaps it does sometimes,
link |
but it's not anywhere close to some fundamental mechanism
link |
that actually makes a thing work.
link |
I don't know if you can sort of comment on that
link |
in terms of using neuroscience
link |
to understand the human mind and neurons.
link |
Are we really far away potentially
link |
from understanding in the way we understand
link |
the transistors enough to be able to build a computer?
link |
So one thing about understanding
link |
is you can understand for fun.
link |
The other one is to understand so you could build things.
link |
And that's when you really have to understand.
link |
In fact, what got me into the field at MIT
link |
was work by David Marr on this very topic.
link |
So David Marr was a professor at MIT,
link |
but he'd done his PhD in neuroscience,
link |
studying just the architectures of the brain.
link |
But he realized that his work, it was on the cerebellum.
link |
He realized that his work, as rigorous as it was,
link |
left him unsatisfied
link |
because he didn't know what the cerebellum was for
link |
and why it had that architecture.
link |
And so he went to MIT and he was in the AI lab there.
link |
And he said, he had this three level approach
link |
that really grabbed my attention.
link |
So when I was an undergrad at UCLA,
link |
I read one of his papers in a class and said,
link |
Because he said, you have to have a computational theory.
link |
What is being computed and why?
link |
An algorithm, how is it being computed?
link |
What are the precise algorithms?
link |
And then the hardware,
link |
how does it get instantiated in the hardware?
link |
And so to really do neuroscience, he argued,
link |
we needed to have understanding at all those levels.
link |
And that really got me.
link |
I loved the neuroscience, but I realized this guy was saying,
link |
if you can't build it, you don't understand it effectively.
link |
And so that's why I went to MIT.
link |
And I had the pleasure of working with David
link |
until he died just a year and a half later.
link |
So there's been that idea that with neuroscience,
link |
we have to have, in some sense, a top down model
link |
of what's being computed and why
link |
that we would then go after.
link |
And the same thing with the, you know,
link |
trying to reverse engineer a computing system
link |
We really need to understand
link |
what the user interface is about
link |
and what are keys on the keyboard for and so forth.
link |
You need to know why to really understand
link |
all the circuitry and what it's for.
link |
Now, we don't, evolution of a natural selection
link |
does not tell us the deeper question that we're asking,
link |
the answer to the deeper question, which is why?
link |
What's this deeper reality and what's it up to and why?
link |
All it tells us is that whatever reality is,
link |
it's not what you see.
link |
What you see is just an adaptive fiction.
link |
So just to linger on this fascinating, bold question
link |
that shakes you out of your dream state.
link |
Does this fiction still help you in building intuitions
link |
as literary fiction does about reality?
link |
The reason we read literary fiction
link |
is it helps us build intuitions and understanding
link |
in indirect ways sneak up to the difficult questions
link |
of human nature, great fiction.
link |
Same with this observed reality.
link |
Does this interface that we get, this fictional interface,
link |
help us build intuition about deeper truths
link |
of how this whole mess works?
link |
Well, I think that each theory that we propose
link |
will give its own answer to that question, right?
link |
So when the physicists are proposing these structures
link |
like the amplituhedron and cosmological polytope,
link |
associahedron and so forth beyond space time,
link |
we can then ask your question for those specific structures
link |
and say, how much information, for example,
link |
does evolution by natural selection
link |
and the kinds of sensory systems that we have right now
link |
give us about this deeper reality?
link |
And why did we evolve this way?
link |
We can try to answer that question from within the deep.
link |
So there's not gonna be a general answer.
link |
I think what we'll have to do is posit
link |
these new deeper theories
link |
and then try to answer your question
link |
within the framework of those deeper theories,
link |
knowing full well that there'll be an even deeper theory.
link |
So is this paralyzing though?
link |
Because how do we know we're not completely adrift
link |
out to sea, lost forever from,
link |
so like that our theories are completely lost.
link |
if we can never truly deeply introspect to the bottom,
link |
if it's always just turtles on top of turtles infinitely,
link |
isn't that paralyzing for the scientific mind?
link |
Well, it's interesting that you say introspect
link |
Because there is one,
link |
again, this isn't the same spirit of what I said before,
link |
which is it depends on what answer you give
link |
to what's beyond space time,
link |
what answer we would give to your question, right?
link |
So, but one answer that is interesting to explore
link |
is something that spiritual traditions have said
link |
for thousands of years, but haven't said precisely.
link |
So we can't take it seriously in science
link |
until it's made precise,
link |
but we might be able to make it precise.
link |
And that is that they've also said something like
link |
space and time aren't fundamental,
link |
they're Maya, they're illusion.
link |
And, but that if you look inside, if you introspect
link |
and let go of all of your particular perceptions,
link |
you will come to something that's beyond conceptual thought.
link |
And that is, they claim,
link |
being in contact with the deep ground of being
link |
that transcends any particular conceptual understanding.
link |
If that is correct, and I'm not saying it's correct,
link |
but, and I'm not saying it's not correct,
link |
I'm just saying, if that's correct,
link |
then it would be the case that as scientists,
link |
because we also are in touch with this ground of being,
link |
we would then not be able
link |
to conceptually understand ourselves all the way,
link |
but we could know ourselves just by being ourselves.
link |
And so we would, there would be a sense
link |
in which there is a fundamental grounding
link |
to the whole enterprise,
link |
because we're not separate from the enterprise.
link |
This is the opposite of the impersonal third person science.
link |
This would make science go personal all the way down.
link |
And, but nevertheless, scientific,
link |
because the scientific method would still be
link |
what we would use all the way down
link |
for the conceptual understanding.
link |
Unfortunately, I still don't know
link |
if you went all the way down.
link |
It's possible that this kind of whatever consciousness is
link |
and we'll talk about it,
link |
is getting the cliche statement of be yourself.
link |
It is somehow digging at a deeper truth of reality,
link |
but you still don't know when you get to the bottom.
link |
A lot of people, they'll take psychedelic drugs
link |
and they'll say, well, that takes my mind to certain places
link |
where it feels like that is revealing
link |
some deeper truth of reality,
link |
but you still, it could be interfaces on top of interfaces.
link |
That's, in your view of this, you really don't know.
link |
I mean, it's Gato's incompleteness
link |
is that you really don't know.
link |
My own view on it, for what it's worth,
link |
because I don't know the right answer,
link |
but my own view on it right now is that it's never ending.
link |
I think that there will never,
link |
that this is great, as I said before,
link |
great job security for science.
link |
And that we, if this is true,
link |
and if consciousness is somehow important
link |
or fundamental in the universe,
link |
this may be an important fundamental fact
link |
about consciousness itself,
link |
that it's a never ending exploration
link |
that's going on in some sense.
link |
Well, that's interesting.
link |
Push back on the job security.
link |
So maybe as we understand this kind of idea
link |
deeper and deeper,
link |
we understand that the pursuit is not a fruitful one.
link |
Then maybe we need to,
link |
maybe that's why we don't see aliens everywhere,
link |
is you get smarter and smarter and smarter,
link |
you realize that exploration is,
link |
there's other fun ways to spend your time than exploring.
link |
You could be sort of living maximally
link |
in some way that's not exploration.
link |
There's all kinds of video games you can construct
link |
and put yourself inside of them
link |
that don't involve you going outside of the game world.
link |
It's a feeling, from my human perspective,
link |
what seems to be fun is challenging yourself
link |
and overcoming those challenges.
link |
So you can constantly artificially generate challenges
link |
for yourself, like Sisyphus and his boulder,
link |
just, and that's it.
link |
So the scientific method
link |
that's always reaching out to the stars,
link |
that's always trying to figure out
link |
the puzzle on the bottom puzzle,
link |
we're always trying to get to the bottom turtle.
link |
Maybe if we can build more and more the intuition
link |
that that's infinite pursuit,
link |
we agree to start deviating from that pursuit
link |
and start enjoying the here and now
link |
versus the looking out into the unknown always.
link |
Maybe that's looking out into the unknown
link |
as a early activity for a species that's evolved.
link |
I'm just sort of saying, pushing back,
link |
as you probably got a lot of scientists excited
link |
in terms of job security,
link |
I could envision where it's not job security,
link |
where scientists become more and more useless.
link |
Maybe they're like the holders of the ancient wisdom
link |
that allows us to study our own history,
link |
but not much more than that, just to push back.
link |
That's good pushback.
link |
I'll put one in there for the scientists again,
link |
but sure, but then I'll take the other side too.
link |
So when Faraday did all of his experiments
link |
with magnets and electricity and so forth,
link |
he came up with all this wonderful empirical data
link |
and James Clerk Maxwell looked at it
link |
and wrote down a few equations,
link |
which we can now write down in a single equation,
link |
the Maxwell equation if we use geometric algebra,
link |
just one equation.
link |
That opened up unbelievable technologies.
link |
People are zooming and talking to each other
link |
around the world, the whole electronics industry.
link |
There was something that transformed our lives
link |
in a very positive way.
link |
With the theories beyond space time,
link |
here's one potential, right now,
link |
most of the galaxies that we see, we can see them,
link |
but we know that we could never get to them
link |
no matter how fast we traveled.
link |
They're going away from us at the speed of light or beyond.
link |
So we can't ever get to them.
link |
So there's all this beautiful real estate
link |
that's just smiling and waving at us
link |
and we can never get to it.
link |
But that's if we go through space time.
link |
But if we recognize that space time
link |
is just a data structure, it's not fundamental.
link |
We're not little things inside space time.
link |
Space time was a little data structure in our perceptions.
link |
It's just the other way around.
link |
Once we understand that,
link |
and we get equations for the stuff that's beyond space time,
link |
maybe we won't have to go through space time.
link |
Maybe we can go around it.
link |
Maybe I can go to Proxima Centauri
link |
and not go through space.
link |
I can just go right there directly.
link |
It's a data structure.
link |
We can start to play with it.
link |
So I think that for what it's worth,
link |
my take would be that the endless sequence of theories
link |
that we could contemplate building
link |
will lead to an endless sequence of new remarkable insights
link |
into the potentialities, the possibilities
link |
that would seem miraculous to us.
link |
And that we will be motivated to continue the exploration
link |
partly just for the technological innovations
link |
But the other thing that you mentioned though,
link |
what about just being?
link |
What if we decide instead of all this doing and exploring,
link |
My guess is that the best scientists will do both
link |
and that the act of being will be a place
link |
where they get many of their ideas
link |
and that they then pull into the conceptual realm.
link |
And I think many of the best scientists,
link |
like Einstein comes to mind, right?
link |
Where these guys say, look,
link |
I didn't come up with these ideas by a conceptual analysis.
link |
I was thinking in vague images
link |
and it was just something nonconceptual.
link |
And then it took me a long, long time
link |
to pull it out into concepts
link |
and then longer to put it into math.
link |
But the real insights didn't come from data.
link |
The real insights didn't come from just slavishly
link |
playing with equations.
link |
They came from a deeper place.
link |
And so there may be this going back and forth
link |
between the complete nonconceptual
link |
where there's essentially no end to the wisdom
link |
and then conceptual systems
link |
where there's the girdle limits that we have to that.
link |
And that may be, if consciousness is important
link |
and fundamental, that may be what consciousness,
link |
at least part of what consciousness is about
link |
is this discovering itself, discovering its possibilities,
link |
so to speak, and we can talk about what that might mean,
link |
by going from the nonconceptual to the conceptual
link |
and back and forth.
link |
So you get better and better and better at being.
link |
Let me ask you just to linger on the evolutionary,
link |
because you mentioned evolutionary game theory
link |
and that's really where you,
link |
the perspective from which you come
link |
to form the case against reality.
link |
At which point in our evolutionary history
link |
do we start to deviate the most from reality?
link |
Is it way before life even originated on Earth?
link |
Is it in the early development from bacteria and so on?
link |
Or is it when some inklings of what we think of
link |
as intelligence or maybe even complex consciousness
link |
started to emerge?
link |
So where did this deviation,
link |
just like with the interfaces in a computer,
link |
you start with transistors and then you have assembly
link |
and then you have C, C++, then you have Python,
link |
then you have GUIs, all that kind,
link |
you have layers upon layers.
link |
When did we start to deviate?
link |
Well, David Marr, again, my advisor at MIT,
link |
in his book, Vision,
link |
suggested that the more primitive sensory systems
link |
were less realistic, less theoretical,
link |
but that by the time you got to something
link |
as complicated as the humans,
link |
we were actually estimating the true shapes
link |
and distances to objects and so forth.
link |
So his point of view, and I think it was probably,
link |
it's not an uncommon view among my colleagues
link |
that, yeah, the sensory systems of lower creatures
link |
may just not be complicated enough
link |
to give them much, much truth.
link |
But as you get to 86 million neurons,
link |
you can now compute the truth,
link |
or at least the parts of the truth that we need.
link |
When I look at evolutionary game theory,
link |
one of my graduate students, Justin Mark,
link |
did some simulations using genetic algorithms.
link |
So there he was just exploring,
link |
we start off with random organisms,
link |
random sensory genetics and random actions.
link |
And the first generation was unbelievably,
link |
it was a foraging situation.
link |
They were foraging for resources.
link |
Most of them stayed in one place,
link |
didn't do anything important.
link |
But we could then just look at how the genes evolved.
link |
And what we found was,
link |
what he found was that basically you never even saw
link |
the truth organisms even come on the stage.
link |
If they came up, they were gone in one generation,
link |
they just weren't.
link |
So they came and went even just in one generation.
link |
They just are not good enough.
link |
The ones that were just tracking,
link |
their senses just were tracking the fitness payoffs
link |
were far more fit than the truth seekers.
link |
So an answer at one level,
link |
I want to give an answer at a deeper level,
link |
but just with evolutionary game theory,
link |
because my attitude as a scientist is,
link |
I don't believe any of our theories.
link |
I take them very, very seriously.
link |
I study them, I look at their implications,
link |
but none of them are the gospel.
link |
They're just the latest ideas that we have.
link |
And so the reason I study evolutionary game theory
link |
is because that's the best tool we have right now
link |
There is nothing else that competes.
link |
And so as a scientist, it's my responsibility
link |
to take the best tools and see what they mean.
link |
And the same thing the physicists are doing.
link |
They're taking the best tools
link |
and looking at what they entail.
link |
But I think that science now has enough experience
link |
to realize that we should not believe our theories
link |
in the sense that we've now arrived.
link |
In 1890, a lot of physicists thought we'd arrived.
link |
They were discouraging bright young students
link |
from going into physics, because it was all done.
link |
And that's precisely the wrong attitude forever.
link |
It's the wrong attitude forever.
link |
The attitude we should have is a century from now,
link |
they'll be looking at us and laughing
link |
at what we didn't know.
link |
And we just have to assume that that's going to be the case.
link |
Just know that everything that we think
link |
is so brilliant right now, our final theory.
link |
A century from now, they'll look at us
link |
like we look at the physicists of 1890 and go,
link |
how could they have been so dumb?
link |
So I don't want to make that mistake.
link |
So I'm not doctrinaire about any
link |
of our current scientific theories.
link |
I am doctrinaire about this.
link |
We should use the best tools we have right now.
link |
That's what we've got.
link |
And with humility.
link |
Well, so let me ask you about game theory.
link |
I love game theory, evolutionary game theory.
link |
But I'm always suspicious of it, like economics.
link |
When you construct models,
link |
it's too easy to construct things that oversimplify
link |
just because we, our human brains,
link |
enjoy the simplification of constructing a few variables
link |
that somehow represent organisms or represent people
link |
and running a simulation that then allows you
link |
to build up intuition and then it feels really good
link |
because you can get some really deep
link |
and surprising intuitions.
link |
But how do you know your models aren't,
link |
the assumptions underlying your models
link |
aren't some fundamentally flawed?
link |
And because of that,
link |
your conclusions are fundamentally flawed.
link |
So I guess my question is what are the limits
link |
in your use of game theory, evolution game theory,
link |
your experience with it?
link |
What are the limits of game theory?
link |
So I've gotten some pushback from professional colleagues
link |
and friends who have tried to rerun simulations
link |
and try to, the idea that we don't see the truth
link |
is not comfortable and so many of my colleagues
link |
are very interested in trying to show that we're wrong.
link |
And so the idea would be to say that somehow
link |
we did something, as you're suggesting,
link |
maybe something special that wasn't completely general.
link |
We got some little special part of the whole search space
link |
in evolutionary game theory in which this happens to be true
link |
but more generally organisms would evolve
link |
So the best pushback we've gotten is from a team at Yale.
link |
And they suggested that if you use
link |
thousands of payoff functions,
link |
so we in our simulations, we just use a couple,
link |
one or two, because it was our first simulations, right?
link |
So that would be a limit.
link |
We had one or two payoff functions,
link |
we showed the result of those,
link |
at least for the genetic algorithms.
link |
And they said, if you have 20,000 of them,
link |
then we can find these conditions in which
link |
truth seeing organisms would be the ones
link |
that evolved and survived.
link |
And so we looked at their simulations
link |
and it certainly is the case that you can find
link |
special cases in which truth can evolve.
link |
So when I say it's probability zero,
link |
it doesn't mean it can't happen.
link |
It can happen, in fact, it could happen infinitely often.
link |
It's just probability zero.
link |
So probability zero things can happen infinitely often.
link |
When you say probability is zero, you mean probability
link |
To be very, very precise.
link |
So for example, if I have a unit square on the plane
link |
and I use a measure on a probability measure
link |
in which the area of a region is this probability.
link |
Then if I draw a curve in that unit square,
link |
it has measure precisely zero,
link |
precisely not approximately, precisely zero.
link |
And yet it has infinitely many points.
link |
So there's an object that for that probability measure
link |
has probability zero, and yet there's
link |
infinitely many points in it.
link |
So that's what I mean when I say that things
link |
that are probability zero can happen
link |
infinitely often in principle.
link |
Yeah, but infinity, as far as, and I look outside often,
link |
I walk around and I look at people.
link |
I have never seen infinity in real life.
link |
That's an interesting issue.
link |
I've been looking, I've been looking.
link |
I don't notice it, infinitely small or the infinitely big.
link |
And so the tools of mathematics,
link |
you could sort of apply the same kind of criticism
link |
that it is a very convenient interface into our reality.
link |
That's a big debate in mathematics,
link |
the intuitionists versus the ones who take,
link |
for example, the real numbers as real.
link |
And that's a fun discussion.
link |
Nicholas Giesen, a physicist,
link |
has really interesting work recently
link |
on how if you go with intuitionist mathematics,
link |
you could effectively quantize Newton,
link |
and you find that the Newtonian theory
link |
and quantum theory aren't that different
link |
once you go with it.
link |
It's really quite interesting.
link |
So the issue he raises is a very, very deep one,
link |
and one that I think we should take quite seriously,
link |
which is how should we think about the reality
link |
of the contours hierarchy?
link |
Aleph one, aleph two, and all these different infinities
link |
versus just a more algorithmic approach, right?
link |
So where everything's computable,
link |
in some sense, everything's finite,
link |
as big as you want, but nevertheless finite.
link |
So yeah, that ultimately boils down to
link |
whether the world is discrete or continuous
link |
in some general sense.
link |
And again, we can't really know,
link |
but there's just a mind breaking thought,
link |
just common sense reasoning,
link |
that something can happen,
link |
and as yet, probability of it happening is 0%.
link |
That doesn't compute for common sense computer.
link |
This is where you have to be a sharp mathematician
link |
to really, and I'm not.
link |
Sharp is one word.
link |
What I'm saying is common sense computer is,
link |
I mean that in a very kind of,
link |
in a positive sense,
link |
because we've been talking about perception systems
link |
and interfaces, if we are to reason about the world,
link |
we have to use the best interfaces we got.
link |
And I'm not exactly sure that game theory
link |
is the best interface we got for this.
link |
In application of mathematics, tricks and tools
link |
in mathematics, the game theory is the best we got
link |
when we are thinking about the nature of reality
link |
and fitness functions and evolution, period.
link |
Well, that's a fair rejoinder,
link |
and I think that that was the tool that we used.
link |
And if someone says, here's a better mathematical tool
link |
and here's why, this mathematical tool
link |
better captures the essence of Darwin's idea,
link |
John Maynard Smith didn't quite get it
link |
with evolutionary game theory.
link |
There's this thing.
link |
Now there are tools like evolutionary graph theory,
link |
which generalize evolutionary game theory,
link |
and then there's quantum game theory.
link |
So you can use quantum tools like entanglement,
link |
for example, as a resource in games
link |
that change the very nature of the solutions,
link |
the optimal solutions of the game theory.
link |
Well, the work from Yale is really interesting.
link |
It's a really interesting challenge of these ideas
link |
where, okay, if you have a very large number
link |
of fitness functions, or let's say you have
link |
a nearly infinite number of fitness functions
link |
or a growing number of fitness functions,
link |
what kind of interesting things start to emerging
link |
if you are to be an organism?
link |
If to be an organism that adapts means
link |
having to deal with an ensemble of fitness functions.
link |
Right, and so we've actually redone some of our own work
link |
based on theirs, and this is the back and forth
link |
that we expect in science, right?
link |
And what we found was that in their simulations,
link |
they were assuming that you couldn't carve the world
link |
up into objects, and so we said,
link |
well, let's relax that assumption.
link |
Allow organisms to create data structures
link |
that we might call objects,
link |
and an object would be you take,
link |
you would do hierarchical clustering
link |
of your fitness payoff functions,
link |
the ones that have similar shapes.
link |
If you have 20,000 of them, maybe these 50
link |
are all very, very similar,
link |
so I can take all the perception, action, fitness stuff
link |
and make that into a data structure,
link |
and we'll call that a unit or an object.
link |
And as soon as we did that,
link |
then all of their results went away.
link |
It turned out they were the special case
link |
and that the organisms that were allowed
link |
to only see, that were shaped to see only fitness payoffs
link |
were the ones that were.
link |
So the idea is that objects then,
link |
what are objects from an evolutionary point of view?
link |
This bottle, we thought that when I saw a bottle,
link |
it was because I was seeing a true object
link |
that existed whether or not it was perceived.
link |
Evolutionary theories suggest a different interpretation.
link |
I'm seeing a data structure that is encoding
link |
a convenient way of looking at various fitness payoffs.
link |
I can use this for drinking.
link |
I could use it as a weapon, not a very good one.
link |
I could be somewhere with head with it.
link |
If my goal is mating, this is pointless.
link |
So I'm seeing for, what I'm coding here
link |
is all sorts of actions and the payoffs that I could get.
link |
When I pick up an apple,
link |
now I'm getting a different set of actions and payoffs.
link |
When I pick up a rock, I'm getting, so for every object,
link |
what I'm getting is a different set of payoff functions
link |
and with various actions.
link |
And so once you allow that,
link |
then what you find is once again that truth goes extinct
link |
and the organisms that just get an interface
link |
are the ones that win.
link |
But the question, just sneaking up on, this is fascinating.
link |
From where do fitness functions originate?
link |
What gives birth to the fitness functions?
link |
So if there's a giant black box
link |
that just keeps giving you fitness functions,
link |
what are we trying to optimize?
link |
You said that water has different uses than an apple.
link |
So there's these objects.
link |
What are we trying to optimize?
link |
And why is not reality a really good generator
link |
of fitness functions?
link |
So each theory makes its own assumptions and says,
link |
grant me this, then I'll explain that.
link |
So evolutionary game theory says,
link |
grant me fitness payoffs, right?
link |
And grant me strategies with payoffs.
link |
And I can write down the matrix
link |
for this strategy interacts with that strategy.
link |
These are the payoffs that come up.
link |
If you grant me that,
link |
then I can start to explain a lot of things.
link |
Now you can ask for a deeper question like,
link |
okay, how does physics evolve biology
link |
and where do these fitness payoffs come from, right?
link |
Now that's a completely different enterprise.
link |
And of course, evolutionary game theory then
link |
would be not the right tool for that.
link |
It would have to be a deeper tool
link |
that shows where evolutionary game theory comes from.
link |
My own take is that there's gonna be a problem
link |
in doing that because space time isn't fundamental.
link |
It's just a user interface.
link |
And that the distinction that we make
link |
between living and nonliving
link |
is not a fundamental distinction.
link |
It's an artifact of the limits of our interface, right?
link |
So this is a new wrinkle and this is an important wrinkle.
link |
It's so nice to take space and time as fundamental
link |
because if something looks like it's inanimate,
link |
it's inanimate and we can just say it's not living.
link |
Now it's much more complicated.
link |
Certain things are obviously living.
link |
I'm talking with you, I'm obviously interacting
link |
with something that's alive and conscious.
link |
I think we've let go of the word obviously
link |
in this conversation.
link |
I think nothing is obvious.
link |
Nothing's obvious, that's right.
link |
But when we get down to like an ant,
link |
it's obviously living, but I'll say it appears to be living.
link |
But when we get down to a virus, now people wonder
link |
and when we get down to protons,
link |
people say it's not living.
link |
And my attitude is look, I have a user interface.
link |
Interface is there to hide certain aspects of reality
link |
and others to, it's an uneven representation,
link |
Certain things just get completely hidden.
link |
Dark matter and dark energy are most of the energy
link |
and matter that's out there.
link |
Our interface just plain flat out hides them.
link |
The only way we get some hint is because
link |
gravitational things are going wrong within our,
link |
so most things are outside of our interface.
link |
The distinction between living and nonliving
link |
is not fundamental.
link |
It's an artifact of our interface.
link |
So if we really, really want to understand
link |
where evolution comes from,
link |
to answer the question, the deep question you asked,
link |
I think the right way we're gonna have to do that
link |
is to come up with a deeper theory than space time
link |
in which there may not be the notion of time
link |
and show that whatever this dynamics of that deeper theory
link |
is, by the way, I'll talk about how you could have dynamics
link |
without time, but the dynamics of this deeper theory,
link |
when we project it into, in certain ways,
link |
then we do get space time and we get what appears
link |
to be evolution by natural selection.
link |
So I would love to see evolution by natural selection,
link |
nature, red and tooth and claw, people fighting,
link |
animals fighting for resources and the whole bit,
link |
come out of a deeper theory in which perhaps
link |
it's all cooperation, there's no limited resources
link |
and so forth, but as a result of projection,
link |
you get space and time, and as a result of projection,
link |
you get nature, red and tooth and claw,
link |
the appearance of it, but it's all an artifact
link |
I like this idea that the line between living
link |
and nonliving is very important
link |
because that's the thing that would emerge
link |
before you have evolution, the idea of death.
link |
So that seems to be an important component
link |
of natural selection, and if that emerged,
link |
because that's also asking the question,
link |
I guess, that I ask, where do fitness functions come from?
link |
That's like asking the old meaning of life question, right?
link |
It's the why, why, why?
link |
And one of the big underlying whys,
link |
okay, you can start with evolution on Earth,
link |
but without living, without life and death,
link |
without the line between the living and the dead,
link |
you don't have evolution.
link |
So what if underneath it, there's no such thing
link |
as the living and the dead?
link |
There's no, like this concept of an organism, period.
link |
There's a living organism that's defined
link |
by a volume in space time that somehow interacts,
link |
that over time maintains its integrity somehow.
link |
It has some kind of history, it has a wall of some kind.
link |
The outside world, the environment,
link |
and then inside, there's an organism.
link |
So you're defining an organism,
link |
and also you define that organism
link |
by the fact that it can move, and it can become alive,
link |
which you kind of think of as moving,
link |
combined with the fact that it's keeping itself
link |
separate from the environment,
link |
so you can point out that thing is living,
link |
and then it can also die.
link |
That seems to be all very powerful components of space time
link |
that enable you to have something
link |
like natural selection and evolution.
link |
Well, and there's a lot of interesting work,
link |
some of it by collaborators of Carl Friston and others,
link |
where they have Bayes net kind of stuff
link |
that they build on the notion of a Markov blanket.
link |
So you have some states within this network
link |
that are inside the blanket, then you have the blanket,
link |
and then the states outside the blanket.
link |
And the states inside this Markov blanket
link |
are conditionally independent of the states
link |
outside the blanket conditioned on the blanket.
link |
And what they're looking at is that the dynamics inside
link |
of the states inside the Markov blanket
link |
seem to be trying to estimate properties of the outside
link |
and react to them in a way.
link |
So it seems like you're doing probabilistic inferences
link |
in ways that might be able to keep you alive.
link |
So there's interesting work going on in that direction.
link |
But what I'm saying is something slightly different,
link |
and that is, like, when I look at you,
link |
all I see is skin, hair, and eyes, right?
link |
But I know that there's a deeper reality.
link |
I believe that there's a much deeper reality.
link |
There's the whole world of your experiences,
link |
your thoughts, your hopes, your dreams.
link |
In some sense, the face that I see
link |
is just a symbol that I create, right?
link |
And as soon as I look away, I delete that symbol.
link |
But I don't delete you.
link |
I don't delete the conscious experience,
link |
the whole world of your...
link |
So I'm only deleting an interface symbol.
link |
But that interface symbol is a portal, so to speak.
link |
Not a perfect portal, but a genuine portal
link |
into your beliefs, into your conscious experiences.
link |
That's why we can have a conversation.
link |
Your consciousness is genuinely affecting mine,
link |
and mine is genuinely affecting yours,
link |
through these icons, which I create on the fly.
link |
I mean, I create your face.
link |
When I look, I delete it.
link |
I don't create you, your consciousness.
link |
That's there all the time, but I do...
link |
So now, when I look at a cat,
link |
I'm creating something that I still call living,
link |
and I still think is conscious.
link |
When I look at an ant, I create something
link |
that I still would call living, but maybe not conscious.
link |
When I look at something I call a virus,
link |
now I'm not even sure I would call it living.
link |
And when I look at a proton, I would say,
link |
I don't even think it's not alive at all.
link |
It could be that I'm nevertheless interacting
link |
with something that's just as conscious as you.
link |
I'm not saying the proton is conscious.
link |
The face that I'm creating when I look at you,
link |
that face is not conscious.
link |
That face is a data structure in me.
link |
That face is an experience.
link |
It's not an experiencer.
link |
Similarly, a proton is something that I create
link |
when I look or do a collision
link |
in the Large Hadron Collider or something like that.
link |
But what is behind the entity in space time?
link |
So I've got this space time interface,
link |
and I've just got this entity that I call a proton.
link |
What is the reality behind it?
link |
Well, the physicists are finding these big, big structures.
link |
The amplitude hadron, the sociahedron,
link |
cause what's behind those?
link |
Could be consciousness, what I'm playing with.
link |
In which case, when I'm interacting with a proton,
link |
I could be interacting with consciousness.
link |
Again, to be very, very clear,
link |
because it's easy to misunderstand,
link |
I'm not saying a proton is conscious.
link |
Just like I'm not saying your face is conscious.
link |
Your face is a symbol I create and then delete as I look.
link |
So your face is not conscious,
link |
but I know that that face in my interface,
link |
the Lex Friedman face that I create,
link |
is an interface symbol that's a genuine portal
link |
into your consciousness.
link |
The portal is less clear for a cat,
link |
even less clear for an ant.
link |
And by the time we get down to a proton,
link |
the portal is not clear at all.
link |
But that doesn't mean I'm not interacting
link |
with consciousness, it just means my interface gave up.
link |
And there's some deeper reality that we have to go after.
link |
So your question really forces out a big part
link |
of this whole approach that I'm talking about.
link |
So it's this portal and consciousness.
link |
I wonder why you can't,
link |
your portal is not as good to a cat,
link |
to a cat's consciousness than it is to a human.
link |
Does it have to do with the fact that you're human
link |
and just similar organisms, organisms of similar complexity
link |
are able to create portals better to each other?
link |
Or is it just as you get more and more complex,
link |
you get better and better portals?
link |
Well, let me answer one aspect of it
link |
that I'm more confident about,
link |
then I'll speculate on that.
link |
Why is it that the portal is so bad with protons?
link |
Well, and elementary particles more generally.
link |
So quarks, leptons and gluons and so forth.
link |
Well, the reason for that is because those are just
link |
symmetries of space time.
link |
More technically, they're irreducible representations
link |
of the Poincare group of space time.
link |
So they're just literally representations
link |
of the data structure of space time that we're using.
link |
So that's why they're not very much insightful.
link |
They're just almost entirely tied
link |
to the data structure itself.
link |
they're telling you only something about the data structure,
link |
not behind the data structure.
link |
It's only when we get to higher levels
link |
that we're starting to, in some sense,
link |
build portals to what's behind space time.
link |
Yeah, so there's more and more complexity built
link |
on top of the interface of space time with the cat.
link |
So you can actually build a portal, right?
link |
Yeah, this interface of face and hair and so on, skin.
link |
There's some syncing going on between humans though,
link |
where we synced, like you're getting
link |
a pretty good representation of the ideas in my head
link |
and starting to get a foggy view of my memories in my head.
link |
Even though this is the first time we're talking,
link |
you start to project your own memories.
link |
You start to solve like a giant hierarchy of puzzles
link |
about a human, because we're all,
link |
there's a lot of similarities, a lot of it rhymes.
link |
So you start to make a lot of inferences
link |
and you build up this model of a person.
link |
You have a pretty sophisticated model
link |
what's going on underneath.
link |
Again, I just, I wonder if it's possible
link |
to construct these models about each other
link |
and nevertheless be very distant from an underlying reality.
link |
There's a lot of work on this.
link |
So there's some interesting work called signaling games
link |
where they look at how people can coordinate
link |
and come to communicate.
link |
There's some interesting work that was done
link |
by some colleagues and friends of mine,
link |
Louis Narens, Natalia Komarova, and Kimberly Jamieson,
link |
where they were looking at evolving color words.
link |
So you have a circle of colors, the color circle,
link |
and they wanted to see if they could get people to cooperate
link |
and how they carved the color circle up into a circle.
link |
Two units of words.
link |
And so they had a game theoretic kind of thing
link |
that they'd had people do.
link |
And what they found was that when they included,
link |
so most people are trichromats,
link |
you have three kinds of cone photoreceptors,
link |
but there are some, a lot of men,
link |
7% of men are dichromats.
link |
They might be missing the red cone photoreceptor.
link |
They found that the dichromats had an outsized influence
link |
on the final ways that the whole space of colors
link |
was carved up and labels attached.
link |
You needed to be able to include the dichromats
link |
in the conversation.
link |
And so they had a bigger influence
link |
on how you made the boundaries of the language.
link |
And I thought that was a really interesting kind of insight
link |
that there's going to be, again, a game,
link |
perhaps a game where evolutionary or genetic algorithm
link |
kind of thing that goes on in terms of learning
link |
to communicate in ways that are useful.
link |
And so, yeah, you can use game theory to actually explore
link |
that are signaling games.
link |
There's a lot of brilliant work on that.
link |
I'm not doing it, but there's work out there.
link |
So if it's okay, let us tackle once more
link |
and perhaps several more times
link |
after the big topic of consciousness.
link |
Okay, this very beautiful, powerful things
link |
that perhaps is the thing that makes us human, what is it?
link |
What's the role of consciousness in,
link |
let's say even just the thing we've been talking about,
link |
which is the formation of this interface, any kind of ways
link |
you want to kind of start talking about it.
link |
Well, let me say first what most of my colleagues say.
link |
99% are, again, assuming that space time is fundamental,
link |
particles and space time, matter is fundamental,
link |
and most are reductionist.
link |
And so the standard approach to consciousness
link |
is to figure out what complicated systems of matter
link |
with the right functional properties
link |
could possibly lead to the emergence of consciousness.
link |
That's the general idea, right?
link |
So maybe you have to have neurons,
link |
maybe only if you have neurons, but that might not be enough.
link |
They have to certain kinds of complexity
link |
in their organization and their dynamics,
link |
certain kind of network abilities, for example.
link |
So there are those who say, for example,
link |
that consciousness arises from orchestrated collapse
link |
of quantum states of microtubules and neurons, certainly.
link |
So this is Hamroff and Penrose, that's kind of.
link |
So you start with something physical,
link |
a property of quantum states of neurons,
link |
of microtubules and neurons,
link |
and you say that somehow an orchestrated collapse
link |
of those is consciousness or conscious experiences.
link |
Or integrated information theory.
link |
Again, you start with something physical,
link |
and if it has the right kind of functional properties,
link |
it's something they call phi,
link |
with the right kind of integrated information,
link |
then you have consciousness.
link |
Or you can be a panpsychist, Philip Goff, for example,
link |
where you might say, well,
link |
in addition to the particles and space and time,
link |
those particles are not just matter,
link |
they also could have, say, a unit of consciousness.
link |
And so, but once again, you're taking space and time
link |
and particles as fundamental,
link |
and you're adding a new property to them,
link |
say, consciousness, and then you have to talk about how
link |
when a proton and an electron get together
link |
to form hydrogen, then how those consciousnesses
link |
merge to or interact to create the consciousness
link |
of hydrogen and so forth.
link |
There's attention schema theory,
link |
which again, this is how neural network processes
link |
representing to the network itself,
link |
its attentional processes, that could be consciousness.
link |
There's global workspace theory,
link |
and neuronal global workspace theory.
link |
So there's many, many theories of this type.
link |
What's common to all of them is they assume
link |
that space time is fundamental.
link |
They assume that physical processes
link |
and space time is fundamental.
link |
Panpsychism adds consciousness as an additional thing,
link |
it's almost dualist in that regard.
link |
And my attitude is our best science is telling us
link |
that space time is not fundamental.
link |
So why is that important here?
link |
Well, for centuries, deep thinkers thought of earth, air,
link |
fire, and water as the fundamental elements.
link |
It was a reductionist kind of idea.
link |
Nothing was more elemental than those,
link |
and you could sort of build everything up from those.
link |
When we got the periodic table of elements,
link |
we realized that, of course,
link |
we want to study earth, air, fire, and water.
link |
There's combustion science for fire.
link |
There's sciences for all these other things,
link |
water and so forth.
link |
So we're gonna do science with these things,
link |
but fundamental, no, no.
link |
If you're looking for something fundamental,
link |
those are the wrong building blocks.
link |
Earth has many, many different kinds of elements
link |
that project into the one thing that we call earth.
link |
If you don't understand that there's silicon,
link |
that there's iron,
link |
that there's all these different kinds of things
link |
that project into what we call earth,
link |
you're hopelessly lost.
link |
You're not fundamental, you're not gonna get there.
link |
And then after the periodic table,
link |
then we came up with quarks, leptons, and gluons,
link |
the particles of the standard model of physics.
link |
And so we actually now know
link |
that if you really want to get fundamental,
link |
the periodic table isn't it.
link |
It's good for chemistry,
link |
and it's just wonderful for chemistry,
link |
but if you're trying to go deep fundamental,
link |
what is the fundamental science?
link |
You're gonna have to go to quarks, leptons,
link |
and gluons and so forth.
link |
Well, now we've discovered space time itself is doomed.
link |
Quarks, leptons, and gluons
link |
are just irreducible representations
link |
of the symmetries of space time.
link |
So the whole framework
link |
on which consciousness research is being based right now
link |
And for me, these are my friends and colleagues
link |
that are doing this, they're brilliant.
link |
They're absolutely, they're brilliant.
link |
I, my feeling is I'm so sad
link |
that they're stuck with this old framework
link |
because if they weren't stuck with earth, air, fire,
link |
and water, you could actually make progress.
link |
So it doesn't matter how smart you are.
link |
If you start with earth, air, fire, and water,
link |
you're not gonna get anywhere, right?
link |
Can I actually just,
link |
because the word doomed is so interesting,
link |
let me give you some options, multiple choice quiz.
link |
Is space time, we could say is reality
link |
the way we perceive it doomed,
link |
Because doomed just means it could still be right
link |
and we're now ready to go deeper.
link |
So it's not wrong, it's not a complete deviation
link |
from a journey toward the truth.
link |
Right, it's like earth, air, fire, and water is not wrong.
link |
There is earth, air, fire, and water.
link |
That's a useful framework, but it's not fundamental.
link |
Right, well, there's also wrong,
link |
which is they used to believe, as I recently learned,
link |
that George Washington was the president,
link |
the first president of the United States,
link |
was bled to death for something
link |
that could have been easily treated
link |
because it was believed that you can get,
link |
actually, I need to look into this further,
link |
but I guess you get toxins out or demons out.
link |
I don't know what you're getting out
link |
with the bleeding of a person.
link |
So that ended up being wrong,
link |
but widely believed as a medical tool.
link |
So it's also possible that our assumption of space time
link |
is not just doomed, but is wrong.
link |
Well, if we believe that it's fundamental, that's wrong.
link |
But if we believe it's a useful tool, that's right.
link |
But bleeding somebody to death
link |
was believed to be a useful tool.
link |
And that was wrong.
link |
It wasn't just not fundamental.
link |
It was very, I'm sure there's cases
link |
in which bleeding somebody would work,
link |
but it would be a very tiny, tiny, tiny percentage of cases.
link |
So it could be that it's wrong,
link |
like it's a side road that's ultimately leading
link |
to a dead end as opposed to a truck stop or something
link |
that you can get off of.
link |
My feeling is not the dead end kind of thing.
link |
I think that what the physicists are finding
link |
is that there are these structures beyond space time,
link |
but they project back into space time.
link |
And so space time, when they say space time is doomed,
link |
They're saying it's doomed in the sense
link |
that we thought it was fundamental.
link |
It's not fundamental.
link |
It's a useful, absolutely useful and brilliant data structure,
link |
but there are deeper data structures
link |
like cosmological polytope and space time is not fundamental.
link |
What is doomed in the sense that it's wrong
link |
Which is saying space time is fundamental, essentially.
link |
The idea that somehow being smaller in space and time
link |
or space time is a fundamental nature of reality,
link |
that's just wrong.
link |
It turned out to be a useful heuristic
link |
for thermodynamics and so forth.
link |
And in several other places,
link |
reductionism has been very useful,
link |
but that's in some sense an artifact
link |
of how we use our interface.
link |
Yeah, so you're saying size doesn't matter.
link |
Okay, this is very important for me to write down.
link |
Ultimately. Ultimately, right.
link |
It's useful for theories like thermodynamics
link |
and also for understanding brain networks
link |
in terms of individual neurons and neurons
link |
in terms of chemical systems inside cells.
link |
That's all very, very useful,
link |
but the idea that we're getting
link |
to the more fundamental nature of reality, no.
link |
When you get all the way down in that direction,
link |
you get down to the quarks and gluons,
link |
what you realize is what you've gotten down to
link |
is not fundamental reality,
link |
just the irreducible representations of a data structure.
link |
That's all you've gotten down to.
link |
So you're always stuck inside the data structure.
link |
So you seem to be getting closer and closer.
link |
I went from neural networks to neurons,
link |
neurons to chemistry, chemistry to particles,
link |
particles to quarks and gluons.
link |
I'm getting closer and closer to the real.
link |
No, I'm getting closer and closer to the actual structure
link |
of the data structure of space and time,
link |
the irreducible representations.
link |
That's what you're getting closer to,
link |
not to a deeper understanding of what's beyond space time.
link |
We'll also refer, we'll return again
link |
to this question of dynamics
link |
because you keep saying that space time is doomed,
link |
but mostly focusing on the space part of that.
link |
It's very interesting to see why time gets the bad cred too
link |
because how do you have dynamics without time
link |
is the thing I'd love to talk to you a little bit about.
link |
But let us return your brilliant whirlwind overview
link |
of the different theories of consciousness
link |
that are out there.
link |
What is consciousness if outside of space time?
link |
If we think that we want to have a model of consciousness,
link |
we as scientists then have to say,
link |
what do we want to write down?
link |
What kind of mathematical modeling
link |
are we gonna write down, right?
link |
And if you think about it, there's lots of things
link |
that you might want to write down about consciousness.
link |
For all the complicated subject.
link |
So most of my colleagues are saying,
link |
let's start with matter or neurons
link |
and see what properties of matter
link |
could create consciousness.
link |
But I'm saying that that whole thing is out.
link |
Space time is doomed, that whole thing is out.
link |
We need to look at consciousness qua consciousness.
link |
In other words, not as something that arises
link |
in space and time, but perhaps as something
link |
that creates space and time as a data structure.
link |
So what do we want?
link |
And here again, there's no hard and fast rule,
link |
but what you as a scientist have to do
link |
is to pick what you think are the minimal assumptions
link |
that are gonna allow you to boot up a comprehensive theory.
link |
That is the trick.
link |
So what do I want?
link |
So what I chose to do was to have three things.
link |
I said that there are conscious experiences.
link |
Feeling of headache, the smell of garlic,
link |
experiencing the color red.
link |
There are, those are conscious,
link |
so that's the primitive of the theory.
link |
And the reason I want few primitives, why?
link |
Because those are the miracles of the theory, right?
link |
The primitives, the assumptions of the theory
link |
are the things you're not going to explain.
link |
Those are the things you assume.
link |
And those experiences, you particularly mean
link |
there's a subjectiveness to them.
link |
That's the thing when people refer
link |
to the hard problem of consciousness,
link |
is it feels like something to look at the color red, okay.
link |
Exactly right, it feels like something to have a headache
link |
or to feel upset to your stomach.
link |
It feels like something.
link |
And so I'm going to grant that in this theory,
link |
there are experiences and they're fundamental in some sense.
link |
So conscious experience.
link |
So they're not derived from physics.
link |
They're not functional properties of particles.
link |
They are sui generis, they exist.
link |
Just like we assume space time exists.
link |
I'm now saying space time is just a data structure.
link |
It doesn't exist independent of conscious experiences.
link |
Sorry to interrupt once again,
link |
but should we be focusing in your thinking on humans alone?
link |
Or is there something about in relation
link |
to other kinds of organisms that have
link |
a sufficiently high level of complexity?
link |
Or even, or is there some kind of generalization
link |
of the panpsychist idea that all consciousness permeates,
link |
Outside of the usual definition
link |
of what matter is inside space time.
link |
So it's beyond human consciousness.
link |
Human consciousness, from my point of view,
link |
would be one of a countless variety of consciousnesses.
link |
And even within human consciousness,
link |
there's countless variety of consciousnesses within us.
link |
I mean, you have your left and right hemisphere.
link |
And apparently if you split the corpus callosum,
link |
the personality of the left hemisphere
link |
and the religious beliefs of the left hemisphere
link |
can be very different from the right hemisphere.
link |
And their conscious experiences can be disjoint.
link |
One could have one conscious experience.
link |
They can play 20 questions.
link |
The left hemisphere can have an idea in its mind
link |
and the right hemisphere has to guess.
link |
And it might not get it.
link |
So even within you,
link |
there is more than just one consciousness.
link |
It's lots of consciousnesses.
link |
So the general theory of consciousness that I'm after
link |
is not just human consciousness.
link |
It's going to be just consciousness.
link |
And I presume human consciousness is a tiny drop
link |
in the bucket of the infinite variety of consciousnesses.
link |
That said, I should clarify that the black hole
link |
of consciousness is the home cat.
link |
I'm pretty sure cats lack, is the embodiment of evil
link |
and lack all capacity for consciousness or compassion.
link |
So I just want to lay that on the table.
link |
That's the theory I'm working on.
link |
I don't have any good evidence, but it's just an intuition.
link |
It's just a shout out.
link |
Sorry to distract.
link |
So that's the first assumption.
link |
The first assumption, that's right.
link |
The second assumption is that
link |
these experiences have consequences.
link |
So I'm going to say that conscious experiences
link |
can trigger other conscious experiences somehow.
link |
So really in some sense, there's two basic assumptions.
link |
There's some kind of causality.
link |
Is there a chain of causality?
link |
Does this relate to dynamics?
link |
I'll say there's a probabilistic relationship.
link |
So I'm trying to be as nonspecific to begin with
link |
and see where it leads me.
link |
So what I can write down are probability spaces.
link |
So a probability space, which contains
link |
the conscious experiences that this consciousness can have.
link |
So I call this a conscious agent, this technical thing.
link |
Annika Harris and I've talked about this
link |
and she rightly cautions me that people will think
link |
that I'm bringing in a notion of a self or agency
link |
and so forth when I say conscious agent.
link |
So I just want to say that I use the term conscious agent
link |
merely as a technical term.
link |
There is no notion of self in my fundamental definition
link |
of a conscious agent.
link |
There are only experiences and probabilistic relationships
link |
of how they trigger other experiences.
link |
So the agent is the generator of the conscious experience?
link |
The agent is a mathematical structure
link |
that includes a probability measure,
link |
the probability space of a possible conscious experiences
link |
and a Markovian kernel, which describes how
link |
if this agent has certain conscious experiences,
link |
how that will affect the experiences
link |
of other conscious agents, including itself.
link |
But you don't think of that as a self?
link |
No, there is no notion of a self here.
link |
There's no notion of really of an agent.
link |
But is there a locality?
link |
Is there an organism?
link |
So this is, these are conscious units, conscious entities.
link |
But they're distinct in some way
link |
because they have to interact.
link |
Well, so here's the interesting thing.
link |
When we write down the mathematics,
link |
when you have two of these conscious agents interacting,
link |
the pair satisfy a definition of a conscious agent.
link |
So they are a single conscious agent.
link |
So there is one conscious agent.
link |
But it has a nice analytic decomposition
link |
into as many conscious agents as you wish.
link |
So that's a nice interface.
link |
It's a very useful scientific interface.
link |
It's a scale free or if you like a fractal like approach
link |
to it in which we can use the same unit of analysis
link |
at all scales in studying consciousness.
link |
But if I want to talk about,
link |
so there's no notion of learning, memory, problem solving,
link |
intelligence, self, agency.
link |
So none of that is fundamental.
link |
So, and the reason I did that was
link |
because I want to assume as little as possible.
link |
Everything I assume is a miracle in the theory.
link |
It's not something you explain, it's something you assume.
link |
So I have to build networks of conscious agents.
link |
If I want to have a notion of a self,
link |
I have to build a self.
link |
I have to build learning, memory, problem solving,
link |
intelligence and planning, all these different things.
link |
I have to build networks of conscious agents to do that.
link |
It's a trivial theorem that networks of conscious agents
link |
are computationally universal, that's trivial.
link |
So anything that we can do with neural networks
link |
or automata, you can do with networks of conscious agents.
link |
But you can also do more.
link |
The events in the probability space need not be computable.
link |
So the Markovian dynamics is not restricted
link |
to computable functions
link |
because the very events themselves need not be computable.
link |
So this can capture any computable theory.
link |
Anything we can do with neural networks,
link |
we can do with conscious agent networks.
link |
But it leaves open the door for the possibility
link |
of noncomputable interactions between conscious agents.
link |
So if we want a theory of memory, we have to build it.
link |
And there's lots of different ways you could build.
link |
We've actually got a paper,
link |
Chris Fields took the lead on this.
link |
And we have a paper called Conscious Agent Networks
link |
where Chris takes the lead and shows how to use
link |
these networks of conscious agents to build memory
link |
and to build primitive kinds of learning.
link |
But can you provide some intuition
link |
of what conscious networks,
link |
networks of conscious agents helps you?
link |
First of all, what that looks like.
link |
And I don't just mean mathematically.
link |
Of course, maybe that might help build up intuition.
link |
But how that helps us potentially solve
link |
the hard problem of consciousness.
link |
Or is that baked in, that that exists?
link |
Can you solve the hard problem of consciousness,
link |
why it tastes delicious when you eat a delicious ice cream
link |
with networks of conscious agents?
link |
Or is that taken as an assumption?
link |
So the standard way the hard problem is thought of
link |
is we're assuming space and time and particles
link |
or neurons, for example.
link |
These are just physical things that have no consciousness.
link |
And we have to explain how the conscious experience
link |
of the taste of chocolate could emerge from those.
link |
So the typical hard problem of consciousness
link |
is that problem, right?
link |
How do you boot up the taste of chocolate,
link |
the experience of the taste of chocolate from neurons, say,
link |
or the right kind of artificial intelligence circuitry?
link |
How do you boot that up?
link |
That's typically what the hard problem of consciousness
link |
means to researchers.
link |
Notice that I'm changing the problem.
link |
I'm not trying to boot up conscious experiences
link |
from the dynamics of neurons or silicon
link |
or something like that.
link |
I'm saying that that's the wrong problem.
link |
My hard problem would go in the other direction.
link |
If I start with conscious experiences,
link |
how do I build up space and time?
link |
How do I build up what I call the physical world?
link |
How do I build up what we call brains?
link |
Because I'm saying consciousness
link |
is not something that brains do.
link |
Brains are something that consciousness makes up.
link |
It's among the experience,
link |
it's an ephemeral experience in consciousness.
link |
I look inside, so to be very, very clear,
link |
right now, I have no neurons.
link |
If you looked, you would see neurons.
link |
That's a data structure that you would create on the fly,
link |
and it's a very useful one.
link |
As soon as you look away,
link |
you garbage collect that data structure,
link |
just like that Necker cube that I was talking about
link |
on the piece of paper.
link |
When you look, you see a 3D cube you created on the fly.
link |
As soon as you look away, that's gone.
link |
When you say you, you mean a human being scientist.
link |
Right now, that's right.
link |
More generally, it'll be conscious agents,
link |
because as you pointed out,
link |
am I asking for a theory of consciousness
link |
only about humans?
link |
No, it's consciousness,
link |
which human consciousness is just a tiny sliver.
link |
But you are saying that there is,
link |
that's a useful data structure.
link |
How many other data structures are there?
link |
That's why I said you human.
link |
If there's another Earth,
link |
if there's another alien civilization
link |
and doing these kinds of investigations,
link |
would they come up with similar data structures?
link |
What is the space of data structures,
link |
I guess is what I'm asking.
link |
My guess is that if consciousness is fundamental,
link |
consciousness is all there is,
link |
then the only thing that mathematical structure
link |
can be about is possibilities of consciousness.
link |
And that suggests to me
link |
that there could be an infinite variety of consciousnesses,
link |
and a vanishingly small fraction of them
link |
use space time data structures
link |
and the kinds of structures that we use.
link |
There's an infinite variety of data structures.
link |
Now, this is very similar
link |
to something that Max Tegmark has said,
link |
but I want to distinguish it.
link |
He has this level four multiverse idea.
link |
He thinks that mathematics is fundamental.
link |
And so that's the fundamental reality.
link |
And since there's an infinite variety of,
link |
endless variety of mathematical structures,
link |
there's an infinite variety of multiverses in his view.
link |
I'm saying something similar in spirit,
link |
but importantly different.
link |
There's an infinite variety
link |
of mathematical structures, absolutely.
link |
But mathematics isn't the fundamental reality
link |
in this framework.
link |
and mathematics is to consciousness
link |
like bones are to an organism.
link |
You need the bones.
link |
So mathematics is not divorced from consciousness,
link |
but it's not the entirety of consciousness by any means.
link |
And so there's an infinite variety of consciousnesses
link |
and signaling games that consciousnesses could interact via.
link |
And therefore worlds, your common worlds,
link |
data structures that they can use to communicate.
link |
So space and time is just one of an infinite variety.
link |
And so I think that what we'll find is that
link |
as we go outside of our little space time bubble,
link |
we will encounter utterly alien forms
link |
of conscious experience that we may not be able
link |
to really comprehend in the following sense.
link |
If I ask you to imagine a color
link |
that you've never seen before,
link |
does anything happen?
link |
And that's just one color.
link |
I'm asking for just a color.
link |
We actually know, by the way,
link |
that apparently there are women called tetraphams
link |
who have four color receptors, not just three.
link |
And Kimberly Jameson and others who've studied these women
link |
have good evidence that they apparently have
link |
a new dimension of color experience
link |
that the rest of us don't have.
link |
So these women are apparently living in a world of color
link |
that you and I can't even concretely imagine.
link |
No man can imagine them.
link |
And yet they're real color experiences.
link |
And so in that sense, I'm saying,
link |
now take that little baby step,
link |
oh, there are women who have color experiences
link |
that I could never have.
link |
Well, that's shocking.
link |
Now take that infinite.
link |
There are consciousnesses where every aspect
link |
of their experiences is like that new color.
link |
It's something utterly alien to you.
link |
You have nothing like that.
link |
And yet these are all possible varieties
link |
of conscious experience.
link |
And when you say there's a lot of consciousnesses,
link |
as a singular consciousness,
link |
basically the set of possible experiences you can have
link |
in that subjective way,
link |
as opposed to the underlying mechanism.
link |
Because you say that, you know,
link |
having a extra color receptor,
link |
ability to have new experiences
link |
that somehow a different consciousness,
link |
is there a way to see that as all the same consciousness,
link |
the subjectivity itself?
link |
Because when we have two of these conscious agents
link |
interacting, the mathematics,
link |
they actually satisfy the definition of a conscious agent.
link |
So in fact, they are a single conscious agent.
link |
So in fact, one way to think about what I'm saying,
link |
I'm postulating with my colleagues,
link |
Chaiton and Chris and others,
link |
Robert Pretner and so forth.
link |
There is one big conscious agent, infinitely complicated.
link |
But fortunately, we can, for analytic purposes,
link |
break it down all the way to,
link |
in some sense, the simplest conscious agent,
link |
which has one conscious experience, one.
link |
This one agent can experience red 35, and that's it.
link |
That's what it experiences.
link |
You can get all the way down to that.
link |
So you think it's possible that consciousness,
link |
is much more, is fundamental,
link |
or at least much more in the direction of the fundamental
link |
than is space time as we perceive it?
link |
That's the proposal.
link |
And therefore, what I have to do,
link |
in terms of the hard problem of consciousness,
link |
is to show how dynamical systems of conscious agents
link |
could lead to what we call space and time
link |
and neurons and brain activity.
link |
In other words, we have to show how you get space time
link |
and physical objects entirely from a theory
link |
of conscious agents outside of space time,
link |
with the dynamics outside of space time.
link |
So that's, and I can tell you how we plan to do that,
link |
but that's the idea.
link |
Okay, the magic of it, that chocolate is delicious.
link |
So there's a mathematical kind of thing
link |
that we could say here, how it can emerge
link |
within this system of networks of conscious agents,
link |
but is there going to be at the end of the proof
link |
why chocolate is so delicious?
link |
I guess I'm going to ask different kinds of dumb questions
link |
to try to sneak up.
link |
Oh, well, that's the right question, and when I say
link |
that I took conscious experiences as fundamental,
link |
what that means is, in the current version of my theory,
link |
I'm not explaining conscious experiences
link |
where they came from.
link |
That's the miracle, that's one of the miracles.
link |
So I have two miracles in my theory.
link |
There are conscious experiences, like the taste of chocolate,
link |
and that there's a probabilistic relationship.
link |
When certain conscious experiences occur,
link |
others are more likely to occur.
link |
Those are the two miracles that are possible.
link |
Is it possible to get beyond that
link |
and somehow start to chip away
link |
at the miracleness of that miracle,
link |
that chocolate is delicious?
link |
I've got my hands full with what I'm doing right now,
link |
but I can just say at top level how I would think about that.
link |
That would get at this
link |
consciousness without form.
link |
This is really tough, because it's consciousness without form
link |
versus the various forms that consciousness takes
link |
for the experiences that it has.
link |
So when I write down a probability space
link |
for these conscious experiences, I say,
link |
here's a probability space
link |
for the possible conscious experiences, right?
link |
It's just like when I write down a probability space
link |
for an experiment.
link |
Like I'm gonna flip a coin twice, right?
link |
And I want to look at the probabilities of various outcomes.
link |
So I have to write down a probability space.
link |
There could be heads, heads, heads, tails,
link |
tails, heads, tails, tails.
link |
So any class of probability you're told,
link |
write down your probability space.
link |
If you don't write down your probability space,
link |
you can't get started.
link |
So here's my probability space for consciousness.
link |
How do I want to interpret that structure?
link |
The structure is just sitting there.
link |
There's gonna be a dynamics that happens on it, right?
link |
Experiences appear and then they disappear,
link |
just like heads appears and disappears.
link |
So one way to think about that fundamental
link |
probability space is that corresponds
link |
to consciousness without any content.
link |
The infinite consciousness that transcends
link |
any particular content.
link |
Well, do you think of that as a mechanism,
link |
as a thing, like the rules that govern the dynamics
link |
of the thing outside of space time?
link |
Isn't that, if you think consciousness is fundamental,
link |
isn't that essentially getting like,
link |
it is solving the hard problem,
link |
which is like from where does this thing pop up,
link |
which is the mechanism of the thing popping up,
link |
whatever the consciousness is,
link |
the different kinds and so on, that mechanism.
link |
And also, the question I want to ask is how tricky
link |
do you think it is to solve that problem?
link |
You've solved a lot of difficult problems
link |
throughout the history of humanity.
link |
There's probably more problems to solve left
link |
than we've solved by like an infinity.
link |
But along that long journey of intelligent species,
link |
when will we solve this consciousness one?
link |
Which is one way to measure the difficulty of the problem.
link |
So I'll give two answers.
link |
There's one problem I think we can solve,
link |
but we haven't solved yet.
link |
And that is the reverse
link |
of what my colleagues call the hard problem.
link |
The problem of how do you start with conscious experiences
link |
in the way that I've just described them and the dynamics
link |
and build up space and time and brains,
link |
that I think is a tough technical problem,
link |
but it's in principle solvable.
link |
So I think we can solve that.
link |
So we would solve the hard problem,
link |
not by showing how brains create consciousness,
link |
but how networks of conscious agents
link |
create what we call the symbols that we call brains.
link |
But does that allow you to, so that's interesting.
link |
That's an interesting idea.
link |
Consciousness creates the brain,
link |
not the brain creates consciousness.
link |
But does that allow you to build the thing?
link |
My guess is that it will enable unbelievable technologies.
link |
Once, and I'll tell you why.
link |
I think it plugs into the work
link |
that the physicists are doing.
link |
So this theory of consciousness will be even deeper
link |
than the structures that the physicists are finding,
link |
like the amplituhedron.
link |
But the other answer to your question is less positive.
link |
As I said earlier, I think that there is no such thing
link |
as a theory of everything.
link |
So that I think that the theory that my team is working on,
link |
this conscious agent theory, is just a 1.0 theory.
link |
We're using probability spaces and Markovian curls.
link |
I can easily see people now saying,
link |
well, we can do better if we go to category theory.
link |
And we can get a deeper, perhaps more interesting.
link |
And then someone will say,
link |
well, now I'll go to topoi theory.
link |
So I imagine that there'll be conscious agents,
link |
five, 10, 3 trillion, 0.0, but I think it will never end.
link |
I think ultimately this question
link |
that we sort of put our fingers on of,
link |
how does the formless give birth to form,
link |
to the wonderful taste of chocolate?
link |
I think that we will always go deeper and deeper,
link |
but we will never solve that.
link |
That in some sense, that will be a primitive.
link |
Maybe it's just the limits of my current imagination.
link |
So I'll just say my imagination right now
link |
doesn't peer that deep.
link |
By the way, I'm saying this,
link |
I don't want to discourage some brilliant 20 year old
link |
who then later on proves me dead wrong.
link |
I hope to be proven dead wrong.
link |
Just like you said, essentially from now,
link |
everything we're saying now, everything you're saying,
link |
all your theories will be laughing stock.
link |
They will respect the puzzle solving abilities
link |
and how much we were able to do with so little.
link |
But outside of that, you will all be just,
link |
the silliness will be entertainment for a teenager.
link |
Especially the silliness when we thought
link |
that we were so smart and we knew it all.
link |
So it would be interesting to explore your ideas
link |
by contrasting, you mentioned Annika, Annika Harris,
link |
you mentioned Philip Goff.
link |
So outside of, if you're not allowed to say
link |
the fundamental disagreement is the fact
link |
that space time is fundamental.
link |
What are interesting distinctions
link |
between ideas of consciousness
link |
between you and Annika, for example?
link |
You guys have, you've been on a podcast together,
link |
I'm sure in private you guys
link |
have some incredible conversations.
link |
So where are some interesting sticking points,
link |
some interesting disagreements,
link |
let's say with Annika first.
link |
Maybe there'll be a few other people.
link |
Well, Annika and I just had a conversation this morning
link |
where we were talking about our ideas
link |
and what we discovered really in our conversation
link |
was that we're pretty much on the same page.
link |
It was really just about consciousness.
link |
Our ideas about consciousness
link |
are pretty much on the same page.
link |
She rightly has cautioned me to,
link |
when I talk about conscious agents,
link |
to point out that the notion of agency
link |
is not fundamental in my theory.
link |
The notion of self is not fundamental
link |
and that's absolutely true.
link |
I can use this network of conscious agents,
link |
I now use as a technical term,
link |
conscious agent is a technical term
link |
for that probability space with the Markovian dynamics.
link |
I can use that to build models of a self
link |
and to build models of agency,
link |
but they're not fundamental.
link |
So she has really been very helpful
link |
in helping me to be a little bit clear about these ideas
link |
and not say things that are misleading.
link |
This is the interesting thing about language, actually,
link |
is that language, quite obviously,
link |
is an interface to truth.
link |
It's so fascinating that individual words
link |
can have so much ambiguity
link |
and the specific choices of a word
link |
within a particular sentence,
link |
within the context of a sentence,
link |
can have such a difference in meaning.
link |
It's quite fascinating,
link |
especially when you're talking about topics
link |
like consciousness, because it's a very loaded term.
link |
It means a lot of things to a lot of people
link |
and the entire concept is shrouded in mystery.
link |
So a combination of the fact that it's a loaded term
link |
and that there's a lot of mystery,
link |
people can just interpret it in all kinds of ways.
link |
And so you have to be both precise
link |
and help them avoid getting stuck
link |
on some kind of side road of miscommunication,
link |
lost in translation because you used the wrong word.
link |
That's interesting.
link |
I mean, because for a lot of people,
link |
consciousness is ultimately connected to a self.
link |
I mean, that's our experience of consciousness
link |
is very, it's connected to this ego.
link |
I mean, I just, I mean, what else could it possibly be?
link |
I can't even, how do you begin to comprehend,
link |
to visualize, to conceptualize a consciousness
link |
that's not connected to like this particular organism?
link |
I have a way of thinking about this whole problem now
link |
that comes out of this framework that's different.
link |
So we can imagine a dynamics of consciousness,
link |
not in space and time, just abstractly.
link |
It could be cooperative for all we know.
link |
It could be very friendly, I don't know.
link |
And you can set up a dynamics, a Markovian dynamics
link |
that is so called stationary.
link |
And that's a technical term,
link |
which means that the entropy effectively is not increasing.
link |
There is some entropy, but it's constant.
link |
So there's no increasing entropy.
link |
And in that sense, the dynamics is timeless.
link |
There is no entropic time, but it's a trivial theorem,
link |
three line proof that if you have a stationary
link |
Markovian dynamics, any projection that you make
link |
of that dynamics by conditional probability.
link |
And if you want, I can state a little bit more,
link |
even more mathematically precisely
link |
for some readers or listeners.
link |
But if any projection you take by conditional probability,
link |
the induced image of that Markov chain
link |
will have increasing entropy.
link |
You will have entropic time.
link |
So I'll be very, very precise.
link |
I have a Markov chain X1, X2 through Xn
link |
where Xn, n goes to infinity, right?
link |
The entropy H, capital H of Xn is equal to the entropy H
link |
of Xn minus one for all n.
link |
So the entropy is the same.
link |
But it's a theorem that H of Xn,
link |
say given X sub one is greater than or equal to
link |
H of Xn minus one given X1.
link |
Sure, where does the greater come from?
link |
Because with the theorem, the three line proof,
link |
H of Xn given X1 is greater than or equal to H of Xn
link |
given X1 and X2 because conditioning reduces.
link |
But then H of Xn minus one given X1, X2
link |
is equal to H of Xn given X2,
link |
Xn minus one given X2 by the Markov property.
link |
And then because it's stationary, it's equal to H of X.
link |
I have to write it down.
link |
Anyway, there's a three line proof.
link |
Sure, but the assumption of stationarity,
link |
we're using a lot of terms that people won't understand,
link |
So there's some kind of, some Markovian dynamics
link |
is basically trying to model some kind of system
link |
with some probabilities and there's agents
link |
and they interact in some kind of way
link |
and you can say something about that system
link |
as it evolves stationarity.
link |
So a stationary system is one that has certain properties
link |
in terms of entropy, very well.
link |
But we don't know if it's stationary or not.
link |
We don't know what the properties.
link |
So you have to kind of take assumptions
link |
and see, okay, well, what does the system behave like
link |
under these different properties?
link |
The more constraints, the more assumptions you take,
link |
the more interesting, powerful things you can say,
link |
but sometimes they're limiting.
link |
That said, we're talking about consciousness here.
link |
How does that, you said cooperative, okay, competitive.
link |
It just, I like chocolate.
link |
I'm sitting here, I have a brain, I'm wearing a suit.
link |
It sure as hell feels like I'm myself.
link |
Now, what, am I tuning in?
link |
Am I plugging into something?
link |
Am I a projection, a simple, trivial projection
link |
into space time from some much larger organism
link |
that I can't possibly comprehend?
link |
How the hell, you're saying some,
link |
you're building up mathematical intuitions, fine, great.
link |
But I'm just, I'm having an existential crisis here
link |
and I'm gonna die soon.
link |
Well, I'll die pretty quickly.
link |
So I wanna figure out why chocolate's so delicious.
link |
So help me out here.
link |
So let's just keep sneaking up to this.
link |
Right, so the whole technical thing was to say this.
link |
Even if the dynamics of consciousness is stationary
link |
so that there is no entropic time,
link |
any projection of it, any view of it
link |
will have the artifact of entropic time.
link |
That's a limited resource.
link |
Limited resources, so that the fundamental dynamics
link |
may have no limits, limited resources whatsoever.
link |
Any projection will have certainly time
link |
as a limited resource
link |
and probably lots of other limited resources.
link |
Hence, we could get competition and evolution
link |
and nature, red and tooth and claw
link |
as an artifact of a deeper system
link |
in which those aren't fundamental.
link |
And in fact, I take it as something
link |
that this theory must do at some point
link |
is to show how networks of conscious agents,
link |
even if they're not resource limited,
link |
give rise to evolution by natural selection
link |
Yeah, but you're saying,
link |
I'm trying to understand how the limited resources
link |
that give rise to,
link |
so first the thing gives rise to time,
link |
that gives rise to limited resource,
link |
that gives rise to evolution by natural selection,
link |
how that has to do with the fact that chocolate's delicious?
link |
Well, it's not gonna do that directly.
link |
It's gonna get to this notion of self.
link |
Oh, it's gonna give you?
link |
The notion of self.
link |
Oh, the evolution gives you the notion of self.
link |
And also of a self separate from other selves.
link |
So the idea would be that.
link |
It's competition, it has life and death,
link |
all those kinds of things.
link |
So it won't, I don't think,
link |
as I said, I don't think that I can tell you
link |
how the formless gives rise
link |
to the experience of chocolate.
link |
Right now, my current theory says
link |
that's one of the miracles I'm assuming.
link |
So my theory can't do it.
link |
And the reason my theory can't do it
link |
is because Hoffman's brain can't do it right now.
link |
But the notion of self, yes.
link |
The notion of self can be an artifact
link |
of the projection of it.
link |
So there's one conscious agent.
link |
Because anytime conscious agents interact,
link |
they form a new conscious agent.
link |
So there's one conscious agent.
link |
Any projection of that one conscious agent
link |
gives rise to time,
link |
even if there wasn't any time in that one conscious agent.
link |
And it gives rise, I want to,
link |
now I haven't proven this.
link |
So this is, so now this is me guessing
link |
where the theory is going to go.
link |
I haven't done this.
link |
There's no paper on this yet.
link |
So now I'm speculating.
link |
My guess is I'll be able to show,
link |
or my brighter colleagues working with me
link |
will be able to show
link |
that we will get evolution of a natural selection,
link |
the notion of individual selves,
link |
individual physical objects and so forth
link |
coming out as a projection of this thing.
link |
And that the self, this then will be really interesting
link |
in terms of how it starts to interact
link |
with certain spiritual traditions, right?
link |
Where they will say that there is a notion of self
link |
that needs to be let go,
link |
which is this finite self
link |
that's competing with other selves
link |
to get more money and prestige and so forth.
link |
That self in some sense has to die.
link |
But there's a deeper self,
link |
which is the timeless being
link |
that precludes, not precludes,
link |
but precedes any particular conscious experiences,
link |
the ground of all experience.
link |
That there's that notion of a deep capital self.
link |
But our little capital, lowercase s selves
link |
could be artifacts of projection.
link |
And it may be that what consciousness is doing
link |
in this framework is, right?
link |
It's projected itself down into a self
link |
that calls itself dawn
link |
and a self that calls itself lax.
link |
And through conversations like this,
link |
it's trying to find out about itself
link |
and eventually transcend the limits
link |
of the dawn and lax little icons that it's using
link |
and that little projection of itself.
link |
Through this conversation,
link |
somehow it's learning about itself.
link |
So that thing dressed me up today
link |
in order to understand itself.
link |
And in some sense, you and I are not separate
link |
from that thing and we're not separate from each other.
link |
Yeah, well, I have to question the fashion choices
link |
All right, so you mentioned you agree
link |
in terms of consciousness on a lot of things with Anika.
link |
Is there somebody, friend or friendly foe
link |
that you disagree with in some nuanced, interesting way
link |
or some major way about consciousness,
link |
about these topics of reality that you return to?
link |
Often, it's like Christopher Hitchens
link |
with Rabbi David Wolpe have had interesting conversations
link |
through years that added to the complexity
link |
and the beauty of their friendship.
link |
Is there somebody like that that over the years
link |
has been a source of disagreement with you
link |
that's strengthened your ideas?
link |
Hmm, my ideas have been really shaped by several things.
link |
One is the physicalist framework
link |
that my scientific colleagues, almost to a person,
link |
have adopted and that I adopted too.
link |
And the reason I walked away from it was
link |
because it became clear that we couldn't start
link |
with unconscious ingredients and boot up consciousness.
link |
Can you define physicalist in contrast to reductionist?
link |
So a physicalist, I would say as someone
link |
who takes space time and the objects within space time
link |
as ontologically fundamental.
link |
Right, and then reductionist is saying
link |
the smaller, the more fundamental.
link |
That's a methodological thing.
link |
That's saying within space time,
link |
as you go to smaller and smaller scales in space,
link |
you get deeper and deeper laws, more and more fundamental
link |
laws and the reduction of temperature
link |
to particle movement was an example of that.
link |
But I think that the reason that worked
link |
was almost an artifact of the nature of our interface.
link |
That was for a long time and your colleagues,
link |
including yourself, were physicalists
link |
and now you broke away.
link |
Broke away because I think you can't start
link |
with unconscious ingredients and boot up consciousness.
link |
And so even with Roger Penrose
link |
where there's like a gray area.
link |
Right, and here's the challenge I would put
link |
to all of my friends and colleagues
link |
who give one specific conscious experience
link |
that you can boot up, right?
link |
So if you think that it's integrated information
link |
and I've asked this of Giulio Tononi a couple times
link |
back in the 90s and then just a couple years ago.
link |
I asked Giulio, okay, so great, integrated information.
link |
So we're all interested in explaining
link |
some specific conscious experiences.
link |
So what is, you know, pick one, the taste of chocolate.
link |
What is the integrated information precise structure
link |
that we need for chocolate and why does that structure
link |
have to be for chocolate and why is it
link |
that it could not possibly be vanilla?
link |
Is there any, I asked him, is there any one specific
link |
conscious experience that you can account for?
link |
Because notice, they've set themselves the task
link |
of booting up conscious experiences from physical systems.
link |
That's the task they've set themselves.
link |
But that doesn't mean they're,
link |
I understand your intuition,
link |
but that doesn't mean they're wrong
link |
just because they can't find a way to boot it up yet.
link |
No, that doesn't mean that they're wrong.
link |
It just means that they haven't done it.
link |
I think it's principled.
link |
The reason is principled,
link |
but I'm happy that they're exploring it.
link |
But the fact is, the remarkable fact is
link |
there's not one theory.
link |
So integrated information theory,
link |
orchestrated collapse of microtubules,
link |
global workspace theory.
link |
These are all theories of consciousness.
link |
These are all theories of consciousness.
link |
There's not a single theory that can give you
link |
a specific conscious experience that they say,
link |
here is the physical dynamics or the physical structure
link |
that must be the taste of chocolate
link |
or whatever one they want.
link |
So you're saying it's impossible.
link |
They're saying it's just hard.
link |
My attitude is, okay, no one said
link |
you had to start with neurons or physical systems
link |
and boot up consciousness.
link |
You guys are just taking that.
link |
You chose that problem.
link |
So since you chose that problem,
link |
how much progress have you made?
link |
Well, when you've not been able to come up
link |
with a single specific conscious experience
link |
and you've had these brilliant people
link |
working on it for decades now,
link |
that's not really good progress.
link |
Let me ask you to play devil's advocate.
link |
Can you try to steel man, steel man meaning
link |
argue the best possible case for reality?
link |
The opposite of your book title.
link |
So, or maybe just stick into consciousness.
link |
Can you take the physicalist view?
link |
Can you steel man the physicalist view
link |
for a brief moment playing devil's advocate too?
link |
Or steel man the person you used to be?
link |
Right, right. She's a physicalist.
link |
What's a good, like saying that you might be wrong
link |
right now, what would be a convincing argument for that?
link |
Well, I think the argument I would give
link |
that I believed was, look,
link |
when you have very simple physical systems,
link |
like a piece of dirt,
link |
there's not much evidence of life or consciousness.
link |
It's only when you get really complicated physical systems
link |
like that have brains and really,
link |
the more complicated the brains,
link |
the more it looks like there's consciousness
link |
and the more complicated that consciousness is.
link |
Surely that means that simple physical systems
link |
don't create much consciousness or if maybe not any,
link |
or maybe panpsychists,
link |
they create the most elementary kinds
link |
of simple conscious experiences,
link |
but you need more complicated physical systems to boot up,
link |
to create more complicated consciousnesses.
link |
I think that's the intuition
link |
that drives most of my colleagues.
link |
And you're saying that this concept of complexity
link |
is ill defined when you ground it to space time.
link |
Oh, I think it's well defined
link |
within the framework of space time, right?
link |
No, it's ill defined relative to what you need
link |
to actually understand consciousness
link |
because you're grounding complexity in just in space time.
link |
Oh, got you, right, right.
link |
Yeah, what I'm saying is if it were true
link |
that space time was fundamental,
link |
then I would have to agree
link |
that if there is such a thing as consciousness,
link |
given the data that we've got,
link |
that complex brains have consciousness and dirt doesn't,
link |
that somehow it's the complexity of the dynamics
link |
or organization, the function of the physical system
link |
that somehow is creating the consciousness.
link |
So under those assumptions, yes,
link |
but when the physicists themselves are telling us
link |
that space time is not fundamental, then I can understand.
link |
See, then the whole picture starts to come into focus.
link |
Why, my colleagues are brilliant, right?
link |
These are really smart people.
link |
I mean, Francis Crick worked on this
link |
for the last 20 years of his life.
link |
These are not stupid people.
link |
These are brilliant, brilliant people.
link |
The fact that we've come up
link |
with not a single specific conscious experience
link |
that we can explain and no hope.
link |
There's no one that says, oh, I'm really close.
link |
I'll have it for you in a year.
link |
No, there's just like, there's this fundamental gap.
link |
So much so that Steve Pinker in one of his writings says,
link |
look, he likes the global workspace theory,
link |
but he says the last dollop of the theory
link |
in which there's something it's like to,
link |
he says, we may have to just stipulate that as a brute fact.
link |
Pinker is brilliant, right?
link |
He understands the state of play
link |
on this problem of the hard problem of consciousness,
link |
starting with physicalist assumptions
link |
and then trying to put up consciousness.
link |
So you've set yourself the problem.
link |
I'm starting with physical stuff
link |
that's not conscious.
link |
I'm trying to get the taste of chocolate out
link |
as maybe some kind of function of the dynamics of that.
link |
We've not been able to do that.
link |
And so Pinker is saying, we may have to punt.
link |
We may have to just stipulate that last bit.
link |
He calls it the last dollop.
link |
And just stipulate it as a bare fact of nature
link |
that there is something it's like.
link |
Well, from my point of view as the physical,
link |
the whole point, the whole promise of the physicalist
link |
was we wouldn't have to stipulate.
link |
I was gonna start with the physical stuff
link |
and explain where the consciousness came from.
link |
If I'm going to stipulate consciousness,
link |
why don't I just stipulate consciousness
link |
and not stipulate all the physical stuff too?
link |
So I'm stipulating less.
link |
I'm saying, okay, I agree.
link |
Which is the panpsychist perspective.
link |
Well, it's actually what I call
link |
the conscious realist perspective.
link |
Panpsychists are effectively dualists, right?
link |
They're saying there's physical stuff
link |
that really is fundamental and then consciousness stuff.
link |
So I would go with Pinker and say, look,
link |
let's just stipulate the consciousness stuff,
link |
but I'm not gonna stipulate the physical stuff.
link |
I'm gonna actually now show how to boot up
link |
the physical stuff from just the consciousness stuff.
link |
So I'll stipulate less.
link |
Is it possible, so if you stipulate less,
link |
is it possible for our limited brains to visualize reality
link |
as we delve deeper and deeper and deeper?
link |
Is it possible to visualize somehow?
link |
With the tools of math, with the tools of computers,
link |
with the tools of our mind, are we hopelessly lost?
link |
You said there's ways to intuit what's true
link |
using mathematics and probability
link |
and sort of a Markovian dynamics, all that kind of stuff,
link |
but that's not visualizing.
link |
That's a kind of building intuition.
link |
But is it possible to visualize
link |
in the way we visualize so nicely in space time
link |
in four dimensions, in three dimensions, sorry.
link |
Well, we really are looking through a two dimensional screen
link |
until it's what we intuit to be a three dimensional world
link |
and also inferring dynamic stuff, making it 4D.
link |
Anyway, is it possible to visualize some pretty pictures
link |
that give us a deeper sense of the truth of reality?
link |
I think that we will incrementally be able to do that.
link |
I think that, for example, the picture that we have
link |
of electrons and photons interacting and scattering,
link |
it may have not been possible
link |
until Faraday did all of his experiments
link |
and then Maxwell wrote down his equations.
link |
And we were then sort of forced by his equations
link |
to think in a new way.
link |
And then when Planck in 1900,
link |
desperate to try to solve the problem
link |
of black body radiation,
link |
what they call the ultraviolet catastrophe
link |
where Newton was predicting infinite energies
link |
where there weren't infinite energies
link |
in black body radiation.
link |
And he in desperation proposed packets of energy.
link |
Then once you've done that,
link |
and then you have an Einstein come along five years later
link |
and show how that explains the photoelectric effect.
link |
And then eventually in 1926, you get quantum theory.
link |
And then you get this whole new way of thinking
link |
that was, from the Newtonian point of view,
link |
completely contradictory and counterintuitive, certainly.
link |
And maybe if Giesen is right, not contradictory.
link |
Maybe if you use intuitionist math, they're not contradictory,
link |
but still, certainly you wouldn't have gone there.
link |
And so here's a case where the experiments
link |
and then a desperate mathematical move,
link |
sort of we use those as a flashlight into the deep fog.
link |
And so that science may be the flashlight into the deep fog.
link |
I wonder if it's still possible to visualize
link |
in the, like we talk about consciousness
link |
from a self perspective experience it.
link |
Hold that idea in our mind,
link |
the way you can experience things directly.
link |
We've evolved to experience things in this 3D world.
link |
And that's a very rich experience.
link |
When you're thinking mathematically,
link |
you still in the end of the day have to project it down
link |
to a low dimensional space to make conclusions.
link |
Your conclusions will be a number or a line
link |
or a plot or a visual.
link |
So I wonder like how we can really touch some deep truth
link |
in a subjective way, like experience it,
link |
really feel the beauty of it, you know,
link |
in the way that humans feel beauty.
link |
Right, are we screwed?
link |
I don't think we're screwed.
link |
I think that we get little hints of it
link |
from psychedelic drugs and so forth.
link |
We get hints that there are certain interventions
link |
that we can take on our interface.
link |
I apply this chemical,
link |
which is just some element of my interface
link |
to this other, to a brain I ingested.
link |
And all of a sudden I seem like I've opened new portals
link |
into conscious experiences.
link |
Well, that's very, very suggestive.
link |
That's like the black body radiation doing something
link |
that we didn't expect, right?
link |
It doesn't go to infinity