The following is a conversation with Leonard Susskind.

He's a professor of theoretical physics at Stanford University and founding director of Stanford Institute of Theoretical Physics.

He is widely regarded as 1 of the fathers of string theory and in general, as 1 of the greatest physicists of our time, both as a researcher and an educator.

This is the Artificial Intelligence Podcast.

Perhaps you noticed that the people I've been speaking with are not just computer scientists, but philosophers, mathematicians, writers, psychologists, physicists, and soon other disciplines.

To me, AI is much bigger than deep learning, bigger than computing.

It is our civilization's journey into understanding the human mind and creating echoes of it in the machine.

If you enjoy the podcast, subscribe on YouTube, give it 5 stars on iTunes, support it on Patreon, or simply connect with me on Twitter at Lex Friedman, spelled F-R-I-D-M-A-N.

And now, here's my conversation with Leonard Susskind.

You worked and were friends with Richard Feynman.

What I saw, I think what I saw was somebody who could do physics in this deeply intuitive way.

His style was almost to close his eyes and visualize the phenomena that he was thinking about.

And through visualization, outflank the mathematical, the highly mathematical and very, very sophisticated technical arguments that people would use.

I think that was also natural to me, but I saw somebody who was actually successful at it, who could do physics in a way that I regarded as simpler, more direct, more intuitive.

And while I don't think he changed my way of thinking, I do think he validated it.

He made me look at it and say, yeah, that's something you can do and get away with.

So do you find yourself, whether you're thinking about quantum mechanics or black holes or string theory, using intuition as a first step or step throughout using visualization?

I tend not to think about the equations, I tend not to think about the symbols, I tend to try to visualize the phenomena themselves.

And then when I get an insight that I think is valid, I might try to convert it to mathematics.

I'm good enough at it, but I'm not a great mathematician.

So for me, the way of thinking about physics is first intuitive, first visualization, scribble a few equations maybe, but then try to convert it to mathematics.

Experience is that other people are better at converting it to mathematics than I am.

And yet you've worked with very counterintuitive ideas.

do you visualize something counterintuitive?

Very little of modern physics is intuitive.

It means the ability to think about it with basic classical physics, the physics that we evolved with throwing stones, splashing water, whatever it happens to be.

Quantum physics, general relativity, quantum field theory are deeply unintuitive in that way.

But, you know, after time and getting familiar with these things, you develop new intuitions.

And it's to the point where me and many of my friends, I and many of my friends, can think more easily quantum mechanically than we can classically.

I mean, yes, our neural wiring in our brain is such that we understand rocks and stones and water and so on.

Do you think it's possible to create a wiring of neuron-like state devices that more naturally understand quantum mechanics, understand wave function, understand these weird things?

Well I'm not sure, I think many of us have evolved the ability to think quantum mechanically to some extent.

But that doesn't mean you can think like an electron.

Just visualizing four-dimensional space or five-dimensional space or six-dimensional space, I think we're fundamentally wired to visualize 3 dimensions.

I can't even visualize 2 dimensions or 1 dimension without thinking about it as embedded in 3 dimensions.

If I want to visualize a line, I think of the line as being a line in 3 dimensions.

Or I think of the line as being a line on a piece of paper with a piece of paper being in 3 dimensions.

I never seem to be able to, in some abstract and pure way, Visualize in my head the 1 dimension the 2 dimension the 4 dimension the 5 dimensions and I don't think that's ever gonna happen The reason is I think neural wiring is just set up for that On the other hand we do learn ways to think about 567 dimensions.

We learn ways, we learn mathematical ways, and we learn ways to visualize them, but they're different.

And so yeah, I think we do rewire ourselves, whether we can ever completely rewire ourselves to be completely comfortable with these concepts, I doubt.

So I'm sure there's somewhat, you could argue, creatures that live in a two-dimensional space.

And while it's romanticizing the notion of course, we're all living as far as we know in three-dimensional space, but how do those creatures imagine 3D space?

Well, probably the way we imagine 4D, by using some mathematics and some equations and some tricks.

Okay, So jumping back to Feynman just for a second, he had a little bit of an ego.

Why, do you think ego is powerful or dangerous in science?

I think you have to have both arrogance and humility.

You have to have the arrogance to say, I can do this.

On the other hand, I think you also have to have the humility to know that you're very likely to be wrong on any given occasion.

Everything you're thinking could suddenly change.

Young people can come along and say things you won't understand and you'll be lost and flabbergasted.

You better recognize that you're very limited and you better be able to say to yourself, I'm not so limited that I can't win this battle with nature.

It takes a special kind of person who can manage both of those, I would say.

And I would say there's echoes of that in your own work.

A little bit of ego, a little bit of outside of the box, humble thinking.

So was there a time where you felt, you looked at yourself and asked, am I completely wrong about this?

Oh yeah, about the whole thing or about specific things?

Me and me and my ability to do this thing.

First of all, did you have those kinds of doubts?

I came from a very working class background and I was uncomfortable in academia for a long time.

But they weren't doubts about my ability.

They were just the discomfort in being in an environment that my family hadn't participated in, I knew nothing about as a young person.

I didn't learn that there was such a thing called physics until I was almost 20 years old.

So I did have certain kind of doubts, but not about my ability.

I don't think I was too worried about whether I would succeed or not.

I never felt this insecurity, am I ever gonna get a job?

That had never occurred to me that I wouldn't.

Maybe you could speak a little bit to this sense of what is academia.

Because I too feel a bit uncomfortable in it.

There's something I can't put quite into words, what you have, that's not, doesn't, if we call it music, you play a different kind of music than a lot of academia.

I don't know that I thought about it as much as I just felt it.

Thinking is 1 thing, feeling is another thing.

I felt like an outsider until a certain age when I suddenly found myself the ultimate insider in academic physics.

And that was a sharp transition, and I wasn't a young man.

So you were never quite, it was a phase transition, you were never quite in the middle.

I always felt a little bit of an outsider.

My approach to mathematics was different.

But also, My social background that I came from was different.

Now these days, half the young people I meet there are parents or professors.

So yeah, But then all of a sudden, at some point, I found myself at the very much the center of, maybe not the only 1 at the center, but certainly 1 of the people in the center of a certain kind of physics.

So, maybe a little bit with Feynman, but in general, how do you develop ideas?

The younger time, I spent more time with myself.

Now, because I'm at Stanford, because I have a lot of ex-students and people who are interested in the same thing I am, I spend a good deal of time, almost on a daily basis, interacting, brainstorming, as you said.

I spend less time probably completely self-focused and with a piece of paper and just sitting there staring at it.

What are your hopes for quantum computers?

So machines that are based on, that have some elements of leveraged quantum mechanical ideas.

It's not just leveraging quantum mechanical ideas.

You can simulate quantum systems on a classical computer.

Simulate them means solve the Schrodinger equation for them or solve the equations of quantum mechanics on a computer, on a classical computer.

But the classical computer is not doing, is not a quantum mechanical system itself.

Of course it is, everything's made of quantum mechanics, but it's not functioning.

It's not functioning as a quantum system.

The quantum computer is truly a quantum system which is actually doing the things that you're programming it to do.

You want to program a quantum field theory?

If you do it in classical physics, that program is not actually functioning in the computer as a quantum field theory.

Physically, it's not doing the things that the quantum system would do.

The quantum computer is really a quantum mechanical system which is actually carrying out the quantum operations.

It intrinsically satisfies the uncertainty principle.

It is limited in the same way that quantum systems are limited by uncertainty and so forth.

That means that what you're doing when you program something for a quantum system is you're actually building a real version of the system.

The limits of a classical computer, Classical computers are enormously limited when it comes to quantum systems.

They're enormously limited because, you've probably heard this before, but in order to store the amount of information that's in a quantum state of 400 spins, that's not very many, 400 I can put in my pocket, 400 pennies in my pocket.

To be able to simulate the quantum state of 400 elementary quantum systems, qubits we call them, to do that would take more information than can possibly be stored in the entire universe if it were packed so tightly that you couldn't pack any more in.

On the other hand, if your quantum computer is composed of 400 qubits, it can do everything 400 qubits can do.

What kind of space, if you just intuitively think about the space of algorithms that that unlocks for us.

So there's a whole complexity theory around classical computers measuring the running time of things, and P, so on.

What kind of algorithms, just intuitively, do you think it unlocks for us?

Okay, so we know that there are a handful of algorithms that can seriously beat classical computers and which can have exponentially more power.

Nobody's exhibited this in the laboratory.

We know that's true, but it also seems more and more that the number of such things is very limited.

Only very, very special problems Exhibit that much advantage for a quantum computer of standard problems To my mind as far as I can tell the great power of quantum computers will actually be to simulate quantum systems If you're interested in a certain quantum system and it's too hard to simulate classically, you simply build a version of the same system.

You build a version of it, you build a model of it that's actually functioning as the system, you run it, and then you do the same thing you would do to the quantum system, you make measurements on it.

The advantages, you can run it much slower.

Why not just do experiments on the real system?

You can't slow them down so that you can poke into them.

You can't modify them in arbitrary kinds of ways to see what would happen if I change the system a little bit.

So I think that quantum computers will be extremely valuable in understanding quantum systems.

At the lowest level of the fundamental laws.

They're actually satisfying the same laws as the systems that they're simulating.

Okay, so on the 1 hand you have things like factoring.

Factoring is the great thing of quantum computers, factoring large numbers.

That doesn't seem that much to do with quantum mechanics.

It seems to be almost a fluke that a quantum computer can solve the factoring problem in a short time.

And those problems seem to be extremely special, rare, and it's not clear to me that there's gonna be a lot of them.

On the other hand, there are a lot of quantum systems, chemistry, there's solid state physics, there's material science, there's quantum gravity, there's all kinds of quantum field theory.

And some of these are actually turning out to be applied sciences as well as very fundamental sciences.

So we probably will run out of the ability to solve equations for these things.

Solve equations by the standard methods of pencil and paper.

Solve the equations by the method of classical computers.

And so what we'll do is we'll build versions of these systems, run them, and run them under controlled circumstances where we can change them, manipulate them, make measurements on them, and find out all the things we wanna know.

So in finding out the things we want to know about very small systems, right?

Is there something we can also find out about the macro level?

About something about the function, forgive me, of our brain, biological systems?

The stuff that's about 1 meter in size versus much, much smaller.

Well, all the excitement is about, among the people that I interact with, is understanding black holes.

There is another kind of quantum system that is big.

And 1 of the things we've learned is that the physics of large quantum computers is in some ways similar to the physics of large quantum black holes.

Now you asked, you didn't ask about quantum computers or systems, you didn't ask about black holes, you asked about brains.

Yeah, about stuff that's in the middle of the 2.

So black holes are, there's something fundamental about black holes that feels to be very different than the brain.

Yes, and they also function in a very quantum mechanical way.

Okay, It is, first of all, unclear to me, but of course it's unclear to me.

I don't even have very many friends who are neuroscientists.

I would like to have more friends who are neuroscientists.

Among the few neuroscientists I've ever talked about about this, they are pretty convinced that the brain functions classically, that it is not intrinsically a quantum mechanical system, or it doesn't make use of the special features, entanglement, coherent superposition.

I sort of hope they're wrong, just because I like the romantic idea that the brain is a quantum system.

The other thing, big systems can be composed of lots of little systems.

Materials, The materials that we work with and so forth are to be large systems, a large piece of material, but they're made out of quantum systems.

Now, 1 of the things that's been happening over the last good number of years is we're discovering materials and quantum systems which function much more quantum mechanically than we imagine.

Topological insulators, this kind of thing, that kind of thing.

Those are macroscopic systems, but they're just superconductors.

Superconductors have a lot of quantum mechanics in them.

You can have a large chunk of superconductor, so it's a big piece of material.

On the other hand, it's functioning and its properties depend very, very strongly on quantum mechanics.

And to analyze them, you need the tools of quantum mechanics.

If we can go on to black holes, and looking at the universe as a information processing system, as a computer, as a giant computer.

What's the power of thinking of the universe as an information processing system?

Or what is, perhaps its use, besides the mathematical use of discussing black holes and your famous debates and ideas around that, to human beings, or life in general, as information processing systems.

But all systems are information processing systems.

You poke them, they change a little bit, they evolve.

All systems are information processing systems.

It certainly feels, consciousness intelligence feels like magic.

If we look at information processing, what are the emergent phenomena that come from viewing the world as an information processing system?

If you ask me about physics, my thoughts may be worth something.

If you ask me about this, I'm not sure my thoughts are worth anything.

But as I said earlier, I think when we do introspection, when we imagine doing introspection and try to figure out what it is when we do and we're thinking, I think we get it wrong.

Everything I've heard about the way the brain functions is so counterintuitive.

For example, you have neurons which detect vertical lines.

You have different neurons which detect lines at 45 degrees.

I never imagined that there were whole circuits which were devoted to vertical lines in the brain.

Doesn't seem to be the way my brain works.

My brain seems to work if I put my finger up vertically or if I put it horizontally or if I put it this way or that way, it seems to me it's the same circuits that are...

The way the brain is compartmentalized seems to be very, very different than what I would have imagined if I were just doing psychological introspection about how things work.

My conclusion is that we won't get it right that way.

I think Maybe computer scientists will get it right eventually.

I don't think there are anywheres near it.

I don't even think they're thinking about it.

But by computer, eventually we will build machines, perhaps, which are complicated enough, And partly engineered, partly evolved, maybe evolved by machine learning and so forth, as machine learning is very interesting.

By machine learning, we will evolve systems and we may start to discover mechanisms that have implications for how we think and for what this consciousness thing is all about.

And we'll be able to do experiments on them and perhaps answer questions that we can't possibly answer by introspection.

In many cases, if you look at even the string theory, when you first think about a system, it seems really complicated, like the human brain.

And through some basic reasoning, in trying to discover fundamental, low-level behavior of the system, you find out that it's actually much simpler.

Do you, 1, have you, is that generally the process?

And 2, do you have that also hope for biological systems as well, for all the kinds of stuff we're studying at the human level.

Of course, physics always begins by trying to find the simplest version of something and analyze it.

Yeah, I mean, there are lots of examples where physics has taken very complicated systems, analyzed them, and found simplicity in them for sure.

I said superconductors before, it's an obvious 1.

A superconductor seems like a monstrously complicated thing with all sorts of crazy electrical properties, magnetic properties, and so forth.

And when it finally is boiled down to its simplest elements, it's a very simple quantum mechanical phenomenon called spontaneous symmetry breaking.

And which we, in other contexts, we learned about and we're very familiar with.

So yeah, I mean, yes, we do take complicated things, make them simple, but what we don't want to do is take things which are intrinsically complicated and fool ourselves into thinking that we can make them simple.

We don't wanna make, I don't know who said this, but we don't wanna make them simpler than they really are.

Is the brain a thing which ultimately functions by some simple rules, or is it just complicated?

In terms of artificial intelligence, nobody really knows what are the limits of our current approaches, you mentioned machine learning, how do we create human level intelligence?

It seems that there's a lot of very smart physicists who perhaps oversimplify the nature of intelligence and think of it as information processing, and therefore there doesn't seem to be any theoretical reason why we can't artificially create human level or superhuman level intelligence.

In fact, the reasoning goes, if you create human level intelligence, the same approach you just used to create human level intelligence should allow you to create superhuman level intelligence very easily, exponentially.

So what do you think that way of thinking that comes from physicists is all about?

I wish I knew, but there's a particular reason why I wish I knew.

I am the senior academic advisor to a group of machine learning physicists.

Now that sounds crazy because I know nothing about the subject.

On the other hand, I'm good at giving advice, so I give them advice on things.

Anyway, I see these young physicists who are approaching the machine learning problem.

See all Lex Fridman transcripts on Youtube

Leonard Susskind: Quantum Mechanics, String Theory and Black Holes | Lex Fridman Podcast #41