Is Consciousness Emergent?

Human & Consciousness
3 followers
1 Video View·Oct 26, 2025
Neil deGrasse Tyson and guest David Krakauer unpack the concept of emergence—how simple parts give rise to complex behaviors and new rules. From gases forming fluids to bird flocks, termite mounds, and even AI, they explore when systems deserve to be called “emergent,” why new languages of description are needed, and how true intelligence involves problem-solving, not memorization.
So let's blow open the topic now then of emergence, because that's a topic of great fascination,
especially in biology, but to the extent that that can apply to other systems,
that would be amazing to get some insight into what's going on on the frontier where
we don't understand what's going on. And just because you and Neil and Gary probably too,
can you please define emergence? Because I hear people use the term and I think a lot of times
they're not using it correctly. So can you please just tell us what is emergence?
That's a completely fair observation. I spend most of my life in horror. So, okay,
first of all, I want to say it's a difficult concept and I think it would be completely
accurate to say that we still have huge amounts of work to do to understand formally what it
means. But I'm now going to, having said all that little caveat, I'll tell you where I think it is.
So let's just start with physics because it's easy. So we just talked about it. We talked about gases,
like the kinetic theory of gases. Put loads of those particles together, right, and you can get
solids at the right temperatures and pressures and you can get fluids.
And it turns out that the mathematical equations you use to describe those two systems are
different, right? And the dimension, the simplicity, if you like, the number of terms
in those equations are different. And so we like fluid dynamics because it's actually quite an
elegant way of describing the behaviour of loads and loads of particles in a particular
temperature and pressure. That is emergence. The fact that you have two things, a new state of
matter with properties that wouldn't really seem to apply at the individual particle level,
and it has a new language, a new language of description and a new language of prediction.
Those are the two hallmarks. Now, it's interesting, you were talking about psychology,
advertising and neuroscience. And that's a beautiful example. So let's imagine, right,
that in order to be a really good psychologist or marketer, you had to be a great neuroscientist.
Of course, a really good marketer doesn't need to be a good neuroscientist because all of that detail
is a little bit like a particle in a gas relative to a fluid. A good marketer is doing something
like fluid dynamics by analogy. They're understanding collective properties that
have their own language, right? And I think a lot of kind of pseudoscience, to be honest,
is where a level that has its own perfectly adequate language starts using the more
reductionist language to give it legitimacy. So emergence, one, new states or phases of
matter or organisation. Two, new languages and descriptions, typically mathematical,
it doesn't have to be, right? And three, the tricky one is not everything deserves to be
called emergent. And so finding that emergent level is actually part of the challenge.
And a lot of, we would argue a lot of economic theory would be better off being replaced by
psychology because the language it's derived or invented doesn't really work. It's not like
fluid dynamics. So the failures of emergence are also really interesting.
Cool, man. So when you get to the state of emergence, what is the prediction accuracy
for you being able to say this is most likely going to be happening here?
Yeah, that's actually the criterion, right? If you'll allow me, I'm going to use another analogy.
Okay. So let's say that I'm proving a mathematical theorem, okay? The way that works,
you write down your axioms, right? Your assumptions, you put down your equations,
and then you have a bunch of, kind of a toolkit for doing deduction. You apply calculus or group
theory, whatever you like, and out pops an answer, right? And the correctness of that answer has
nothing to do with your psychological state, nothing to do with what you have for breakfast
in the morning, nothing to do with the economy, nothing to do with neurons. It has to do with the
formal logic of mathematics. And that's the most beautiful example of emergence for me because
you wouldn't gain additional insight into the correctness of a proof by knowing what the
neurons were doing. They're irrelevant. And the technical term for that is screened off.
Truly emergent phenomena screen off microscopic degrees of freedom. That's a fancy language.
And we know it's true, right? But we know it's true of mathematics. It's probably the best
example, I think, of a truly emergent language. And because, to your point, it does predict,
it does deduce. You do get to the right answer, right? You don't have to go down.
So I'm curious about something here. So when we look at flocking birds,
that is a macroscopic group behavior where, as far as I understand it, you cannot,
there's no known way to analyze a single bird in any way that will tell you that in the company of
other birds it will flock. And I've always thought of emergence as just such a system.
And you hinted to that with the gas particle and the gas as a fluid or as a solid or as a gas.
The molecule as a solid, liquid, or gas. So would you agree that these other kinds of systems,
you can't look at a termite and say, one day it will build a termite mound. Is it because we don't
know enough about it? Do we need to be more reductive or less reductive in our analysis
of the organism to know what it would do macroscopically in a group?
Yeah, no, that's really interesting. So I think sometimes that and sometimes the opposite. So let
me give you an example. No, it's really interesting because let's say, it's sort of
interesting, right? If you said, I know everything about the neuroscience of an ant, right? Or a
starling, you know? I know about fluid dynamics, hydrodynamics, I know what feathers do, I know
how far they can see, you know, all of that. So I could predict if I put a bunch of them together,
how they would behave. And I think that there are going to be cases where that is true.
But that doesn't mean emergence isn't still useful. Because you might say, yes, I can.
I need a deep thought. The computer from, you know, Hitchhiker's Guide to the Galaxy, right?
To work it out. And it took the lifetime of the university to do so. But I could do it,
as opposed to, you know what, Neil, I've got a pencil and paper here. I'm going to write down
my little emergent theory, and I'm going to do it in five seconds. And so there is a cytomemergence,
which is just about efficiency. So we're going to see an equation on a t-shirt sometime soon.
We have many of those. Unfortunately, we have too many.
All right. Well, I mean, Chuck mentioned AI earlier on, and we don't want to roll it down
the hill because it's too expensive. Absolutely.
Can you predict the emergence of consciousness in something like an AI?
Wow. Or define the complexity of
life itself in this context. Oh, you want to go even deeper. Fair enough.
Yeah, I mean, why not? Well, yeah, I mean, honestly, and wouldn't that kind of be the
same? Because if an AI really does have emergent consciousness and truly emergent intelligence,
then it really is us at that point. It's just us at that point. It's us in a different form.
Well, let's find out. Okay, let's find out. I might annoy you now.
So the first thing I'm going to say is that AIs have no intelligence.
Okay. And then we'll discuss what that means. But they have tons of capability.
And I'll tell you the difference. I mean, here's my thought experiment. I was asked these kind of
zealots of the technocratic era, which is the following. You have two students. Okay.
And let's call them A and B. And you set them the same exam. It's just a general knowledge quiz.
Right. And they come back. We'll get all the answers right. All right. And I said to you,
which is the better student? You'd say, I don't know. They got the same answer. Now I say to you,
you know, A did the exam in the library, where every time a question came up, they looked up
the answer. And B actually took the exam, I don't know, by the side of the ocean, you know. And you
said, well, clearly B is the better. Now the problem is, so knowledgeable, we know the difference
between fake knowledgeable and real knowledgeable, because we can ask, did you do it in a library or
not? And the problem with a lot of AI at the moment is, it's basically fake intelligence,
as far as I'm concerned. It's a very quick lookup. It's really just, it's very quick,
essentially a really clever lookup. It's a plus. And I'm not saying it's not an amazing technology.
I'm just saying, but it's a very capable technology. If you ask, and again, now we get
to intelligence, kind of my field, what is intelligence? Intelligence is basically someone
or something that makes a hard problem easy. If you went to school and you're sitting down trying
to work out a problem, you look over at the person next to you and they've made a problem
look effortless, you'd say, oh my God, that's pretty intelligent.
No, no, no. You're the intelligent one looking over the shoulder,
because you got the answer easier than they did.
Yeah, that's true. That's like a strategic intelligence.
If I may echo your story with an example I give often, where let's say I'm an architect and I'm
a summer intern and they're both the same on paper. Therefore, they get to come in for an
interview. And I want to pick one from the interview. And this is a contrived example,
but I think you'll agree, David, that there's a church steeple outside my window. And I say,
just for grins, how tall is that church steeple? And the person says, oh, it's 135 feet.
I will say, well, how do you know? I memorized all the church steeple heights. It's a thing I do.
The other person says, I don't know. I'll be right back. It goes away for 10 minutes,
then comes back and says, somewhere between 130 and 140 feet. I said, well, how did you find out?
Said, well, I know how tall I am. And I measured my shadow. Then I measured the
shadow of the church steeple. And then I did some simple math to get this answer.
Who are you going to hire? I'm actually going to hire the
second guy. The second one.
Because clearly what he was able to do was problem solve.
Oh, I said he. I'm sorry. I apologize. Because they were able to problem solve.
And does my example resonate with you here? Yeah, very much so. And I think you can see
where, notice, right, that in the era of the Turing test, the imitation game, you just ask
how high is the steeple. And if it gives you the right answer, you say, look, there you go.
It's indistinguishable from another human being. But you then went a bit further and asked for an
explanation, right? Tell me how you arrived at that answer. Prove to me you understand.
And I think these ideas of it, understanding and explanation are really important to intelligence
and under discussed in place of what Alan Turing did. And, you know, he's my compatriot. I love him.
But the idea of the Turing test did a lot of harm because it allows for this possibility,
essentially, of cheating. I had not thought about it that way, but you've just convinced me
because that's the litmus test that has always been applied. And then everyone is left thinking
there's intelligence on the other side. But David just unpacked that and lays it bare for what it
is. However, in defense of Turing, I think it was just the wrong terminology because basically
the idea was you wouldn't know the difference. That was the test. You would not know the
difference. Correct. Correct. So he's not necessarily saying that they're the same.
He's just saying that one is represented in a way that is indistinguishable from the other.
Which you went further and said, how did you figure this out?
Well, now you just screwed him.
Oh.
