Leonard Susskind: Quantum Mechanics, String Theory and Black Holes

00:00:00.000 | The following is a conversation with Leonard Susskind.

00:00:03.080 | He's a professor of theoretical physics

00:00:04.840 | at Stanford University and founding director

00:00:07.640 | of Stanford Institute of Theoretical Physics.

00:00:10.440 | He is widely regarded as one of the fathers

00:00:13.040 | of string theory and in general,

00:00:14.920 | as one of the greatest physicists of our time,

00:00:17.400 | both as a researcher and an educator.

00:00:20.400 | This is the Artificial Intelligence Podcast.

00:00:23.320 | Perhaps you noticed that the people I've been speaking with

00:00:26.120 | are not just computer scientists,

00:00:27.880 | but philosophers, mathematicians, writers, psychologists,

00:00:31.240 | physicists, and soon other disciplines.

00:00:34.160 | To me, AI is much bigger than deep learning,

00:00:37.360 | bigger than computing.

00:00:38.960 | It is our civilization's journey

00:00:40.840 | into understanding the human mind

00:00:42.840 | and creating echoes of it in the machine.

00:00:45.800 | If you enjoy the podcast, subscribe on YouTube,

00:00:48.760 | give it five stars on iTunes, support it on Patreon,

00:00:51.800 | or simply connect with me on Twitter

00:00:53.640 | at Lex Friedman, spelled F-R-I-D-M-A-N.

00:00:57.560 | And now, here's my conversation with Leonard Susskind.

00:01:01.640 | You worked and were friends with Richard Feynman.

00:01:05.320 | How has he influenced you,

00:01:07.160 | changed you as a physicist and thinker?

00:01:09.120 | - What I saw, I think what I saw was somebody

00:01:13.880 | who could do physics in this deeply intuitive way.

00:01:18.560 | His style was almost to close his eyes

00:01:21.080 | and visualize the phenomena that he was thinking about

00:01:24.520 | and through visualization,

00:01:26.840 | outflank the mathematical, the highly mathematical

00:01:32.080 | and very, very sophisticated technical arguments

00:01:35.480 | that people would use.

00:01:37.040 | I think that was also natural to me,

00:01:39.400 | but I saw somebody who was actually successful at it,

00:01:43.960 | who could do physics in a way that I regarded

00:01:47.760 | as simpler, more direct,

00:01:53.360 | more intuitive, and while I don't think

00:01:56.800 | he changed my way of thinking,

00:01:59.040 | I do think he validated it.

00:02:01.480 | He made me look at it and say,

00:02:02.640 | "Yeah, that's something you can do and get away with."

00:02:06.280 | Practically, you can get away with it.

00:02:08.680 | - So, do you find yourself,

00:02:11.240 | whether you're thinking about quantum mechanics

00:02:13.160 | or black holes or string theory,

00:02:15.920 | using intuition as a first step

00:02:19.400 | or a step throughout using visualization?

00:02:22.160 | - Yeah, very much so, very much so.

00:02:24.680 | I tend not to think about the equations.

00:02:27.720 | I tend not to think about the symbols.

00:02:30.180 | I tend to try to visualize the phenomena themselves.

00:02:34.660 | And then when I get an insight that I think is valid,

00:02:38.400 | I might try to convert it to mathematics,

00:02:40.360 | but I'm not a natural mathematician.

00:02:44.400 | I'm good enough at it.

00:02:46.280 | I'm good enough at it, but I'm not a great mathematician.

00:02:49.820 | So for me, the way of thinking about physics

00:02:52.480 | is first intuitive, first visualization,

00:02:56.700 | scribble a few equations maybe,

00:02:59.720 | but then try to convert it to mathematics.

00:03:03.020 | Experience is that other people are better

00:03:04.840 | at converting it to mathematics than I am.

00:03:06.940 | - And yet, you've worked with very counterintuitive ideas.

00:03:11.520 | So how do you-- - No, that's true.

00:03:12.880 | That's true. - So how do you visualize

00:03:14.120 | something counterintuitive?

00:03:15.640 | How do you dare?

00:03:16.480 | - By rewiring your brain in new ways.

00:03:19.940 | Yeah, quantum mechanics is not intuitive.

00:03:22.720 | Very little of modern physics is intuitive.

00:03:25.760 | Intuitive, what does intuitive mean?

00:03:29.500 | It means the ability to think about it

00:03:31.820 | with basic classical physics,

00:03:33.740 | the physics that we evolved with,

00:03:38.740 | throwing stones, splashing water,

00:03:42.100 | whatever it happens to be.

00:03:43.400 | Quantum physics, general relativity,

00:03:47.100 | quantum field theory are deeply unintuitive in that way.

00:03:51.180 | But after time and getting familiar with these things,

00:03:55.220 | you develop new intuitions.

00:03:57.080 | I always said you rewire.

00:03:58.560 | And it's to the point where me and many of my friends,

00:04:04.400 | I and many of my friends,

00:04:05.740 | can think more easily quantum mechanically

00:04:10.260 | than we can classically.

00:04:11.520 | We've gotten so used to it.

00:04:13.520 | - I mean, yes, our neural wiring in our brain

00:04:17.240 | is such that we understand rocks and stones and water

00:04:20.420 | and so on. - We sort of evolved for that.

00:04:22.040 | - Evolved for it. - Yeah.

00:04:23.320 | - Do you think it's possible to create a wiring

00:04:26.740 | of neuron-like state devices that more naturally

00:04:31.160 | understand quantum mechanics, understand wave function,

00:04:35.880 | understand these weird things?

00:04:38.100 | - Well, I'm not sure.

00:04:39.040 | I think many of us have evolved the ability

00:04:42.480 | to think quantum mechanically to some extent,

00:04:46.360 | but that doesn't mean you can think like an electron.

00:04:48.960 | That doesn't mean, another example,

00:04:53.600 | forget for a minute quantum mechanics,

00:04:55.880 | just visualizing four-dimensional space

00:04:58.660 | or five-dimensional space or six-dimensional space,

00:05:02.080 | I think we're fundamentally wired

00:05:05.780 | to visualize three dimensions.

00:05:08.240 | I can't even visualize two dimensions or one dimension

00:05:11.960 | without thinking about it as embedded in three dimensions.

00:05:16.040 | If I want to visualize a line,

00:05:18.280 | I think of the line as being a line in three dimensions.

00:05:21.840 | Or I think of the line as being a line on a piece of paper

00:05:25.580 | with a piece of paper being in three dimensions.

00:05:28.160 | I never seem to be able to, in some abstract and pure way,

00:05:33.160 | visualize in my head the one dimension, the two dimension,

00:05:36.720 | the four dimension, the five dimensions,

00:05:38.800 | and I don't think that's ever gonna happen.

00:05:41.360 | The reason is, I think,

00:05:42.360 | our neural wiring is just set up for that.

00:05:44.880 | On the other hand, we do learn ways

00:05:49.000 | to think about five, six, seven dimensions.

00:05:52.560 | We learn ways, we learn mathematical ways,

00:05:56.160 | and we learn ways to visualize them, but they're different.

00:05:59.600 | And so, yeah, I think we do rewire ourselves.

00:06:04.760 | Whether we can ever completely rewire ourselves

00:06:07.200 | to be completely comfortable with these concepts, I doubt.

00:06:11.660 | - So that it's completely natural.

00:06:13.500 | - To a degree, it's completely natural.

00:06:15.060 | - So I'm sure there's somewhat, you could argue,

00:06:18.140 | creatures that live in a two-dimensional space.

00:06:22.380 | - Yeah, maybe there are.

00:06:23.580 | - And while it's romanticizing the notion,

00:06:28.220 | of course, we're all living, as far as we know,

00:06:30.020 | in three-dimensional space,

00:06:31.660 | but how do those creatures imagine 3D space?

00:06:35.580 | - Well, probably the way we imagine 4D,

00:06:37.580 | by using some mathematics and some equations

00:06:40.440 | and some tricks.

00:06:43.020 | - Okay, so jumping back to Feynman just for a second,

00:06:48.320 | he had a little bit of an ego.

00:06:51.740 | - Yes.

00:06:54.780 | - Why, do you think ego is powerful or dangerous in science?

00:06:59.120 | - I think both, both, both.

00:07:02.860 | I think you have to have both arrogance and humility.

00:07:06.500 | You have to have the arrogance to say, "I can do this.

00:07:10.560 | "Nature is difficult, nature is very, very hard.

00:07:13.560 | "I'm smart enough, I can do it.

00:07:16.060 | "I can win the battle with nature."

00:07:19.000 | On the other hand, I think you also have to have

00:07:21.000 | the humility to know that you're very likely

00:07:26.000 | to be wrong on any given occasion.

00:07:29.040 | Everything you're thinking could suddenly change.

00:07:32.040 | Young people can come along and say things

00:07:35.500 | you won't understand and you'll be lost and flabbergasted.

00:07:38.920 | So I think it's a combination of both.

00:07:42.600 | You better recognize that you're very limited

00:07:45.720 | and you better be able to say to yourself,

00:07:49.200 | "I'm not so limited that I can't win

00:07:51.220 | "this battle with nature."

00:07:52.560 | It takes a special kind of person who can manage

00:07:58.160 | both of those, I would say.

00:07:59.840 | - And I would say there's echoes of that in your own work,

00:08:03.040 | a little bit of ego, a little bit of outside

00:08:05.360 | of the box, humble thinking.

00:08:06.840 | - I hope so.

00:08:09.580 | - So was there a time where you felt,

00:08:14.580 | you looked at yourself and asked,

00:08:18.020 | "Am I completely wrong about this?"

00:08:19.920 | - Oh yeah, about the whole thing or about specific things?

00:08:23.980 | - The whole thing.

00:08:25.020 | Wait, which whole thing?

00:08:27.060 | - Me and me and my ability to do this thing.

00:08:29.980 | - Oh, those kinds of doubts.

00:08:31.420 | First of all, did you have those kinds of doubts?

00:08:33.860 | - No, I had different kind of doubts.

00:08:35.940 | I came from a very working class background

00:08:37.900 | and I was uncomfortable in academia for a long time.

00:08:42.660 | But they weren't doubts about my ability or my,

00:08:46.260 | they were just the discomfort in being in an environment

00:08:52.700 | that my family hadn't participated in,

00:08:56.600 | I knew nothing about as a young person.

00:08:58.300 | I didn't learn that there was such a thing called physics

00:09:00.700 | until I was almost 20 years old.

00:09:02.340 | - Yeah.

00:09:03.180 | - So I did have certain kind of doubts,

00:09:09.580 | but not about my ability.

00:09:11.860 | I don't think I was too worried

00:09:14.000 | about whether I would succeed or not.

00:09:15.900 | I never felt this insecurity, am I ever gonna get a job?

00:09:21.840 | That had never occurred to me that I wouldn't.

00:09:25.740 | - Maybe you could speak a little bit to this sense

00:09:29.860 | of what is academia?

00:09:31.720 | Because I too feel a bit uncomfortable in it.

00:09:35.100 | There's something I can't put quite into words

00:09:40.020 | what you have that's not, doesn't, if we call it music,

00:09:45.020 | you play a different kind of music than a lot of academia.

00:09:48.500 | How have you joined this orchestra?

00:09:51.940 | How do you think about it?

00:09:54.500 | - I don't know that I thought about it

00:09:56.100 | as much as I just felt it.

00:09:59.300 | Thinking is one thing, feeling is another thing.

00:10:02.800 | I felt like an outsider until a certain age

00:10:07.100 | when I suddenly found myself the ultimate insider

00:10:10.740 | in academic physics.

00:10:12.080 | That was a sharp transition.

00:10:18.640 | I wasn't a young man, I was probably 50 years old.

00:10:22.300 | - So you were never quite, it was a phase transition,

00:10:24.960 | you were never quite in the middle.

00:10:27.360 | - Yeah, that's right, I wasn't.

00:10:29.480 | I always felt a little bit of an outsider.

00:10:32.120 | In the beginning, a lot an outsider.

00:10:34.300 | My way of thinking was different.

00:10:40.480 | My approach to mathematics was different.

00:10:43.080 | But also, my social background

00:10:47.200 | that I came from was different.

00:10:49.360 | Now these days, half the young people I meet,

00:10:51.540 | their parents were professors.

00:10:53.080 | That was not my case.

00:10:55.840 | So, yeah, but then all of a sudden,

00:11:00.840 | at some point, I found myself at very much the center of,

00:11:06.240 | maybe not the only one at the center,

00:11:07.840 | but certainly one of the people in the center

00:11:09.840 | of a certain kind of physics.

00:11:11.340 | And all that went away, I mean, it went away in a flash.

00:11:16.100 | - So, maybe a little bit with Feynman,

00:11:21.760 | but in general, how do you develop ideas?

00:11:24.500 | Do you work through ideas alone?

00:11:26.100 | Do you brainstorm with others?

00:11:27.820 | - Oh, both, both, very definitely both.

00:11:30.380 | The younger time, I spent more time with myself.

00:11:35.560 | Now, because I'm at Stanford,

00:11:39.480 | because I have a lot of ex-students,

00:11:44.480 | and people who are interested in the same thing I am,

00:11:52.840 | I spend a good deal of time, almost on a daily basis,

00:11:56.360 | interacting, brainstorming, as you said.

00:11:58.580 | It's a very important part.

00:12:01.360 | I spend less time, probably, completely self-focused

00:12:05.820 | and with a piece of paper

00:12:09.780 | and just sitting there staring at it.

00:12:11.640 | - What are your hopes for quantum computers?

00:12:16.820 | So, machines that are based on,

00:12:19.740 | that have some elements of leveraged quantum

00:12:23.140 | mechanical ideas.

00:12:25.020 | - It's not just leveraging quantum mechanical ideas.

00:12:28.300 | You can simulate quantum systems on a classical computer.

00:12:33.300 | Simulate them means solve the Schrodinger equation for them,

00:12:36.660 | or solve the equations of quantum mechanics

00:12:40.240 | on a computer, on a classical computer.

00:12:43.660 | But the classical computer is not doing,

00:12:47.440 | is not a quantum mechanical system itself.

00:12:49.820 | Of course it is, everything's made of quantum mechanics,

00:12:52.160 | but it's not functioning.

00:12:53.340 | It's not functioning as a quantum system.

00:12:56.140 | It's just solving equations.

00:12:58.600 | The quantum computer is truly a quantum system,

00:13:01.860 | which is actually doing the things

00:13:04.180 | that you're programming it to do.

00:13:07.400 | You want to program a quantum field theory?

00:13:11.200 | If you do it in classical physics,

00:13:13.700 | that program is not actually functioning

00:13:16.740 | in the computer as a quantum field theory.

00:13:18.800 | It's just solving some equations.

00:13:20.500 | Physically, it's not doing the things

00:13:23.760 | that the quantum system would do.

00:13:27.280 | The quantum computer is really a quantum mechanical system,

00:13:30.420 | which is actually carrying out the quantum operations.

00:13:34.140 | You can measure it at the end.

00:13:36.360 | It intrinsically satisfies the uncertainty principle.

00:13:40.520 | It is limited in the same way that quantum systems

00:13:44.340 | are limited by uncertainty and so forth.

00:13:47.460 | And it really is a quantum system.

00:13:49.260 | That means that what you're doing

00:13:51.420 | when you program something for a quantum system

00:13:53.380 | is you're actually building a real version of the system.

00:13:56.940 | The limits of a classical computer,

00:14:00.460 | classical computers are enormously limited

00:14:02.940 | when it comes to quantum systems.

00:14:07.180 | They're enormously limited

00:14:09.060 | because you've probably heard this before,

00:14:12.280 | but in order to store the amount of information

00:14:14.980 | that's in a quantum state of 400 spins,

00:14:19.940 | that's not very many,

00:14:20.980 | 400 I can put in my pocket,

00:14:22.900 | 400 pennies in my pocket.

00:14:25.600 | To be able to simulate the quantum state

00:14:32.220 | of 400 elementary quantum systems,

00:14:35.900 | qubits we call them,

00:14:37.660 | to do that would take more information

00:14:39.900 | than can possibly be stored in the entire universe

00:14:43.160 | if it were packed so tightly

00:14:45.860 | that you couldn't pack any more in.

00:14:50.420 | 400 qubits.

00:14:52.260 | On the other hand,

00:14:53.380 | if your quantum computer is composed of 400 qubits,

00:14:56.360 | it can do everything 400 qubits can do.

00:14:58.640 | - What kind of space,

00:15:00.780 | if you just intuitively think about the space of algorithms

00:15:04.380 | that that unlocks for us.

00:15:06.360 | So there's a whole complexity theory

00:15:08.520 | around classical computers

00:15:10.100 | measuring the running time of things,

00:15:12.460 | and P, so on.

00:15:13.780 | What kind of algorithms, just intuitively,

00:15:15.640 | do you think it unlocks for us?

00:15:18.400 | - Okay, so we know that there are a handful of algorithms

00:15:22.120 | that can seriously beat classical computers

00:15:25.820 | and which can have exponentially more power.

00:15:28.220 | This is a mathematical statement.

00:15:29.460 | Nobody's exhibited this in the laboratory.

00:15:32.340 | It's a mathematical statement.

00:15:33.580 | We know that's true,

00:15:35.460 | but it also seems more and more

00:15:37.660 | that the number of such things is very limited.

00:15:40.380 | Only very, very special problems

00:15:45.100 | exhibit that much advantage for a quantum computer,

00:15:48.780 | of standard problems.

00:15:52.140 | To my mind, as far as I can tell,

00:15:53.940 | the great power of quantum computers

00:15:55.620 | will actually be to simulate quantum systems.

00:15:58.720 | If you're interested in a certain quantum system

00:16:02.800 | and it's too hard to simulate classically,

00:16:05.900 | you simply build a version of the same system.

00:16:09.860 | You build a version of it,

00:16:11.140 | you build a model of it

00:16:12.060 | that's actually functioning as the system,

00:16:14.480 | you run it, and then you do the same thing

00:16:16.900 | you would do to the quantum system,

00:16:18.540 | you make measurements on it,

00:16:20.460 | quantum measurements on it.

00:16:21.860 | The advantages, you can run it much slower.

00:16:26.180 | You could say, why bother?

00:16:27.540 | Why not just use the real system?

00:16:29.620 | Why not just do experiments on the real system?

00:16:32.380 | Well, real systems are kind of limited.

00:16:33.940 | You can't change them, you can't manipulate them,

00:16:36.340 | you can't slow them down so that you can poke into them.

00:16:39.440 | You can't modify them in arbitrary kinds of ways

00:16:43.460 | to see what would happen

00:16:45.280 | if I change the system a little bit.

00:16:48.780 | I think that quantum computers

00:16:52.100 | will be extremely valuable

00:16:55.580 | in understanding quantum systems.

00:17:00.980 | - At the lowest level of the fundamental laws.

00:17:04.340 | - They're actually satisfying the same laws

00:17:06.540 | as the systems that they're simulating.

00:17:09.180 | - That's right.

00:17:10.020 | - Okay, so on the one hand, you have things like factoring.

00:17:13.060 | Factoring is the great thing of quantum computers,

00:17:17.580 | factoring large numbers.

00:17:19.740 | That doesn't seem that much to do with quantum mechanics.

00:17:22.620 | It seems to be almost a fluke

00:17:24.220 | that a quantum computer can solve

00:17:30.120 | the factoring problem in a short time.

00:17:32.920 | And those problems seem to be extremely special, rare,

00:17:38.520 | and it's not clear to me

00:17:40.300 | that there's gonna be a lot of them.

00:17:42.740 | On the other hand, there are a lot of quantum systems.

00:17:45.180 | Chemistry, there's solid state physics,

00:17:47.900 | there's material science, there's quantum gravity,

00:17:51.020 | there's all kinds of quantum field theory.

00:17:53.640 | And some of these are actually turning out

00:17:56.780 | to be applied sciences

00:17:58.100 | as well as very fundamental sciences.

00:18:00.140 | So we probably will run out of the ability

00:18:05.140 | to solve equations for these things.

00:18:07.820 | You know, solve equations by the standard methods

00:18:09.940 | of pencil and paper,

00:18:11.620 | solve the equations by the method of classical computers.

00:18:15.440 | And so what we'll do is we'll build versions

00:18:18.760 | of these systems, run them,

00:18:22.100 | and run them under controlled circumstances

00:18:24.260 | where we can change them, manipulate them,

00:18:26.840 | make measurements on them,

00:18:28.100 | and find out all the things we wanna know.

00:18:30.540 | - So in finding out the things we wanna know

00:18:33.660 | about very small systems, right,

00:18:36.700 | is there something we can also find out

00:18:40.260 | about the macro level?

00:18:42.100 | About something about the function,

00:18:44.380 | forgive me, of our brain, biological systems?

00:18:48.100 | The stuff that's about one meter in size

00:18:50.380 | versus much, much smaller.

00:18:52.440 | - Well, all the excitement is about,

00:18:55.220 | among the people that I interact with,

00:18:56.880 | is understanding black holes.

00:18:58.920 | - Black holes.

00:18:59.760 | - Black holes are big things.

00:19:01.020 | They are many, many degrees of freedom.

00:19:04.120 | There is another kind of quantum system that is big.

00:19:08.720 | It's a large quantum computer.

00:19:10.860 | And one of the things we've learned

00:19:13.160 | is that the physics of large quantum computers

00:19:15.720 | is in some ways similar to the physics

00:19:17.580 | of large quantum black holes.

00:19:19.760 | And we're using that relationship.

00:19:22.000 | Now you asked, you didn't ask about quantum computers

00:19:24.760 | or systems, you didn't ask about black holes,

00:19:28.120 | you asked about brains.

00:19:30.000 | - Yeah, about stuff that's in the middle of the two.

00:19:32.320 | - It's different.

00:19:33.280 | - So black holes are, there's something fundamental

00:19:38.640 | about black holes that feels to be very different

00:19:41.280 | than brains.

00:19:42.240 | - Yes, and they also function

00:19:44.400 | in a very quantum mechanical way.

00:19:46.000 | - Right.

00:19:46.840 | - Okay, it is, first of all, unclear to me,

00:19:50.660 | but of course it's unclear to me.

00:19:52.720 | I'm not a neuroscientist.

00:19:54.500 | I have, I don't even have very many friends

00:19:58.120 | who are neuroscientists.

00:20:00.240 | I would like to have more friends who are neuroscientists.

00:20:02.640 | I just don't run into them very often.

00:20:04.540 | Among the few neuroscientists I've ever talked about

00:20:08.440 | about this, they are pretty convinced

00:20:12.480 | that the brain functions classically.

00:20:15.760 | That it is not intrinsically a quantum mechanical system

00:20:20.640 | or doesn't make use of the special features,

00:20:23.680 | entanglement, coherent superposition.

00:20:26.480 | Are they right?

00:20:27.440 | I don't know.

00:20:29.000 | I sort of hope they're wrong,

00:20:30.360 | just because I like the romantic idea

00:20:32.960 | that the brain is a quantum system.

00:20:35.240 | But I think probably not.

00:20:38.720 | The other thing, big systems can be composed

00:20:41.640 | of lots of little systems.

00:20:43.040 | Materials, the materials that we work with and so forth

00:20:47.720 | are, can be large systems, a large piece of material,

00:20:52.720 | but they're made out of quantum systems.

00:20:55.120 | Now, one of the things that's been happening

00:20:57.180 | over the last good number of years

00:21:00.600 | is we're discovering materials and quantum systems

00:21:04.720 | which function much more quantum mechanically

00:21:08.200 | than we imagine.

00:21:09.640 | Topological insulators, this kind of thing,

00:21:12.040 | that kind of thing.

00:21:13.520 | Those are macroscopic systems,

00:21:15.220 | but they're just superconductors.

00:21:17.880 | Superconductors have a lot of quantum mechanics in them.

00:21:21.900 | You can have a large chunk of superconductor,

00:21:25.040 | so it's a big piece of material.

00:21:26.800 | On the other hand, it's functioning and its properties

00:21:29.640 | depend very, very strongly on quantum mechanics.

00:21:32.840 | And to analyze them, you need the tools of quantum mechanics.

00:21:36.220 | - If we can go on to black holes,

00:21:39.620 | and looking at the universe as a information processing

00:21:44.780 | system, as a computer, as a giant computer.

00:21:46.760 | - It's a giant computer.

00:21:48.600 | - What's the power of thinking of the universe

00:21:50.920 | as an information processing system?

00:21:52.360 | Or what is, perhaps its use,

00:21:55.200 | besides the mathematical use of discussing black holes

00:21:58.800 | and your famous debates and ideas around that,

00:22:02.840 | to human beings, or life in general

00:22:07.040 | as information processing systems?

00:22:08.840 | - Well, all systems are information processing systems.

00:22:11.560 | You poke them, they change a little bit,

00:22:15.360 | they evolve.

00:22:16.800 | All systems are information processing systems.

00:22:18.400 | - So there's no extra magic to us humans?

00:22:21.460 | It certainly feels, consciousness and intelligence

00:22:25.080 | feels like magic.

00:22:26.080 | - It sure does.

00:22:26.920 | - Where does it emerge from?

00:22:28.240 | If we look at information processing,

00:22:31.780 | what are the emerging phenomena that come from

00:22:36.560 | viewing the world as an information processing system?

00:22:39.920 | - Here is what I think.

00:22:42.000 | My thoughts are not worth much in this.

00:22:43.560 | If you ask me about physics,

00:22:44.660 | my thoughts may be worth something.

00:22:46.680 | If you ask me about this,

00:22:48.080 | I'm not sure my thoughts are worth anything.

00:22:50.920 | But, as I said earlier,

00:22:53.320 | I think when we do introspection,

00:22:55.800 | when we imagine doing introspection

00:22:57.560 | and try to figure out what it is when we do,

00:22:59.520 | when we're thinking, I think we get it wrong.

00:23:03.520 | I'm pretty sure we get it wrong.

00:23:04.880 | Everything I've heard about the way the brain functions

00:23:07.280 | is so counterintuitive.

00:23:08.880 | For example, you have neurons which detect vertical lines.

00:23:14.400 | You have different neurons which detect lines at 45 degrees.

00:23:17.920 | You have different neurons.

00:23:19.480 | I never imagined that there were whole circuits

00:23:21.840 | which were devoted to vertical lines in the brain.

00:23:25.400 | It doesn't seem to be the way my brain works.

00:23:28.140 | My brain seems to work if I put my finger up vertically,

00:23:31.160 | or if I put it horizontally,

00:23:32.240 | or if I put it this way or that way.

00:23:33.420 | It seems to me it's the same circuits that are,

00:23:36.240 | it's not the way it works.

00:23:37.740 | The way the brain is compartmentalized

00:23:41.520 | seems to be very, very different

00:23:43.720 | than what I would have imagined

00:23:45.600 | if I were just doing psychological introspection

00:23:49.680 | about how things work.

00:23:50.880 | My conclusion is that we won't get it right that way.

00:23:55.560 | How will we get it right?

00:23:58.820 | I think maybe computer scientists will get it right.

00:24:02.720 | Eventually, I don't think there are anywheres near it.

00:24:04.480 | I don't even think they're thinking about it.

00:24:06.720 | But by computer, eventually we will build machines,

00:24:09.840 | perhaps, which are complicated enough,

00:24:13.480 | and partly engineered, partly evolved,

00:24:18.380 | maybe evolved by machine learning and so forth.

00:24:21.080 | This machine learning is very interesting.

00:24:23.520 | By machine learning, we will evolve systems,

00:24:26.040 | and we may start to discover mechanisms

00:24:29.440 | that have implications for how we think

00:24:35.360 | and for what this consciousness thing is all about.

00:24:39.520 | And we'll be able to do experiments on them

00:24:42.040 | and perhaps answer questions

00:24:43.720 | that we can't possibly answer by introspection.

00:24:49.640 | - So that's a really interesting point.

00:24:51.720 | In many cases, if you look at even a string theory,

00:24:55.240 | when you first think about a system,

00:24:56.760 | it seems really complicated, like the human brain.

00:24:59.600 | And through some basic reasoning

00:25:01.480 | and trying to discover fundamental,

00:25:05.400 | low-level behavior of the system,

00:25:07.520 | you find out that it's actually much simpler.

00:25:10.160 | Do you, one, have you, is that generally the process?

00:25:13.560 | And two, do you have that also hope

00:25:15.560 | for biological systems as well,

00:25:17.920 | for all the kinds of stuff we're studying

00:25:20.600 | at the human level?

00:25:21.720 | - Of course, physics always begins by trying

00:25:23.480 | to find the simplest version of something and analyze it.

00:25:26.560 | Yeah, I mean, there are lots of examples

00:25:28.400 | where physics has taken very complicated systems,

00:25:33.360 | analyzed them, and found simplicity in them, for sure.

00:25:37.000 | I said superconductors before.

00:25:38.880 | It's an obvious one.

00:25:39.840 | A superconductor seems like a monstrously complicated thing

00:25:42.480 | with all sorts of crazy electrical properties,

00:25:45.960 | magnetic properties, and so forth.

00:25:48.360 | And when it finally is boiled down

00:25:50.440 | to its simplest elements,

00:25:52.920 | it's a very simple quantum mechanical phenomenon

00:25:56.080 | called spontaneous symmetry breaking,

00:25:58.280 | which we, in other contexts, we learned about

00:26:04.760 | and we're very familiar with.

00:26:06.720 | So yeah, I mean, yes, we do take complicated things,

00:26:10.480 | make them simple, but what we don't wanna do

00:26:13.680 | is take things which are intrinsically complicated

00:26:16.520 | and fool ourselves into thinking

00:26:18.400 | that we can make them simple.

00:26:20.600 | We don't wanna make, I don't know who said this,

00:26:22.360 | but we don't wanna make them simpler

00:26:23.640 | than they really are, okay?

00:26:26.720 | Is the brain a thing which ultimately functions

00:26:30.800 | by some simple rules, or is it just complicated?

00:26:35.560 | In terms of artificial intelligence,

00:26:37.760 | nobody really knows what are the limits

00:26:40.840 | of our current approaches.

00:26:41.720 | You mentioned machine learning.

00:26:43.000 | How do we create human-level intelligence?

00:26:44.840 | It seems that there's a lot of very smart physicists

00:26:48.240 | who perhaps oversimplify the nature of intelligence

00:26:51.240 | and think of it as information processing,

00:26:53.800 | and therefore there doesn't seem to be

00:26:55.200 | any theoretical reason why we can't artificially create

00:26:59.080 | human-level or superhuman-level intelligence.

00:27:03.000 | In fact, the reasoning goes,

00:27:04.560 | if you create human-level intelligence,

00:27:07.320 | the same approach you just used

00:27:08.640 | to create human-level intelligence

00:27:10.440 | should allow you to create superhuman-level intelligence

00:27:13.760 | very easily, exponentially.

00:27:16.000 | So what do you think that way of thinking

00:27:18.960 | that comes from physicists is all about?

00:27:22.240 | - I wish I knew, but there's a particular reason

00:27:24.160 | why I wish I knew.

00:27:25.160 | I have a second job.

00:27:29.360 | I consult for Google.

00:27:32.500 | Not for Google, for Google X.

00:27:34.840 | I am the senior academic advisor

00:27:37.880 | to a group of machine learning physicists.

00:27:42.100 | Now, that sounds crazy

00:27:44.760 | because I know nothing about the subject.

00:27:46.800 | I know very little about the subject.

00:27:50.000 | On the other hand, I'm good at giving advice,

00:27:52.240 | so I give them advice on things.

00:27:53.680 | Anyway, I see these young physicists

00:27:56.280 | who are approaching the machine learning problem.

00:27:58.660 | There is a real machine learning problem,

00:28:00.880 | namely, why does it work as well as it does?

00:28:03.080 | Nobody really seems to understand

00:28:06.320 | why it is capable of doing the kind of generalizations

00:28:09.480 | that it does and so forth.

00:28:11.640 | And there are three groups of people

00:28:15.000 | who have thought about this.

00:28:17.480 | There are the engineers.

00:28:19.080 | The engineers are incredibly smart,

00:28:21.580 | but they tend not to think as hard

00:28:23.800 | about why the thing is working

00:28:26.080 | as much as they do how to use it.

00:28:28.240 | Obviously, they've provided a lot of data.

00:28:31.800 | And it is they who demonstrated

00:28:34.040 | that machine learning can work much better

00:28:35.820 | than you had any right to expect.

00:28:37.400 | The machine learning systems are systems,

00:28:40.160 | the systems not too different

00:28:41.960 | than the kind of systems that physicists study.

00:28:44.900 | There's not all that much difference

00:28:46.740 | between quantum, in the structure of the mathematics,

00:28:51.320 | physically, yes, but in the structure of the mathematics,

00:28:54.480 | between a tensor network designed

00:28:57.680 | to describe a quantum system on the one hand,

00:29:01.420 | and the kind of networks that are used in machine learning.

00:29:05.120 | So, more and more, I think, young physicists

00:29:10.040 | are being drawn to this field of machine learning,

00:29:12.800 | some very, very good ones.

00:29:15.120 | I work with a number of very good ones,

00:29:16.800 | not on machine learning, but on having lunch.

00:29:20.360 | - On having lunch?

00:29:21.320 | - Right. (laughs)

00:29:22.440 | - Yeah. (laughs)

00:29:23.840 | - And I can tell you, they are super smart.

00:29:27.600 | They don't seem to be so arrogant

00:29:30.560 | about their physics backgrounds

00:29:32.040 | that they think they can do things that nobody else can do.

00:29:35.080 | But the physics way of thinking, I think,

00:29:37.480 | will add great value to,

00:29:40.880 | will bring value to the machine learning.

00:29:43.960 | I believe it will.

00:29:45.760 | And I think it already has.

00:29:47.120 | - At what time scale do you think predicting

00:29:51.040 | the future becomes useless?

00:29:53.180 | In your long experience in being surprised

00:29:56.520 | at new discoveries.

00:29:57.660 | - Sometimes a day, sometimes 20 years.

00:30:03.440 | There are things which I thought

00:30:05.560 | we were very far from understanding,

00:30:09.800 | which practically in a snap of the fingers

00:30:12.380 | or a blink of the eye suddenly became understood,

00:30:17.380 | completely surprising to me.

00:30:19.040 | There are other things which I looked at

00:30:22.960 | and I said, "We're not gonna understand these things

00:30:25.560 | "for 500 years, in particular quantum gravity."

00:30:29.240 | The scale for that was 20 years, 25 years.

00:30:32.840 | And we understand a lot,

00:30:33.760 | and we don't understand it completely now by any means,

00:30:35.880 | but I thought it was 500 years to make any progress.

00:30:40.740 | It turned out to be very, very far from that.

00:30:42.920 | It turned out to be more like 20 or 25 years

00:30:45.160 | from the time when I thought it was 500 years.

00:30:48.340 | - So if we may, can we jump around quantum gravity,

00:30:51.900 | some basic ideas in physics?

00:30:53.640 | What is the dream of string theory, mathematically?

00:30:59.300 | What is the hope?

00:31:00.140 | Where does it come from?

00:31:01.440 | What problem is it trying to solve?

00:31:03.560 | - I don't think the dream of string theory

00:31:05.000 | is any different than the dream

00:31:06.540 | of fundamental theoretical physics altogether.

00:31:09.560 | - Understanding a unified theory of everything.

00:31:12.680 | - I don't like thinking of string theory

00:31:15.040 | as a subject unto itself,

00:31:17.320 | with people called string theorists

00:31:19.440 | who are the practitioners of this thing

00:31:22.240 | called string theory.

00:31:24.120 | I much prefer to think of them as theoretical physicists

00:31:28.180 | trying to answer deep fundamental questions about nature,

00:31:32.080 | in particular gravity,

00:31:33.400 | in particular gravity and its connection

00:31:35.040 | with quantum mechanics,

00:31:36.320 | and who at the present time

00:31:40.080 | find string theory a useful tool,

00:31:42.540 | rather than saying there's this subject

00:31:45.440 | called string theorists.

00:31:46.440 | I don't like being referred to as a string theorist.

00:31:48.680 | - Yes, but as a tool,

00:31:50.880 | is it useful to think about our nature

00:31:54.120 | in multiple dimensions, the strings vibrating?

00:31:57.520 | - I believe it is useful.

00:31:59.080 | I'll tell you what the main use of it has been up till now.

00:32:02.160 | Well, it has had a number of main uses.

00:32:03.960 | Originally, string theory was invented,

00:32:06.400 | and I know that I was there,

00:32:07.560 | I was right at the spot where it was being invented,

00:32:10.320 | literally, and it was being invented

00:32:15.280 | to understand hadrons.

00:32:17.000 | Hadrons are subnuclear particles,

00:32:19.160 | protons, neutrons, mesons,

00:32:20.960 | and at that time, the late '60s, early '70s,

00:32:27.000 | it was clear from experiment

00:32:30.160 | that these particles called hadrons could vibrate,

00:32:33.720 | could rotate, could do all the things

00:32:36.840 | that a little closed string can do,

00:32:39.680 | and it was and is a valid and correct theory

00:32:46.680 | of these hadrons.

00:32:47.880 | It's been experimentally tested,

00:32:50.120 | and that is a done deal.

00:32:52.440 | It had a second life as a theory of gravity,

00:32:56.200 | the same basic mathematics,

00:32:58.060 | except on a very, very much smaller distance scale.

00:33:01.280 | The objects of gravitation are 19 orders of magnitude

00:33:07.300 | smaller than a proton,

00:33:10.080 | but the same mathematics turned up.

00:33:12.000 | The same mathematics turned up.

00:33:14.300 | What has been its value?

00:33:15.980 | Its value is that it's mathematically rigorous in many ways

00:33:20.760 | and enabled us to find mathematical structures

00:33:25.760 | which have both quantum mechanics and gravity with rigor.

00:33:31.080 | We can test out ideas.

00:33:34.060 | We can test out ideas.

00:33:35.140 | We can't test them in the laboratory,

00:33:37.220 | that they're 19 orders of magnitude too small,

00:33:39.840 | the things that we're interested in,

00:33:41.280 | but we can test them out mathematically

00:33:44.200 | and analyze their internal consistency.

00:33:46.840 | By now, 40 years ago, 35 years ago, and so forth,

00:33:52.780 | people very, very much questioned the consistency

00:33:57.100 | between gravity and quantum mechanics.

00:33:59.240 | Stephen Hawking was very famous for it, rightly so.

00:34:02.840 | Now, nobody questions that consistency anymore.

00:34:09.080 | They don't because we have mathematically precise

00:34:12.440 | string theories which contain both gravity

00:34:15.920 | and quantum mechanics in a consistent way.

00:34:18.940 | So it's provided that certainty

00:34:23.640 | that quantum mechanics and gravity can coexist.

00:34:26.060 | That's not a small thing.

00:34:27.440 | It's a very big thing. - It's a huge thing.

00:34:28.360 | - It's a huge thing.

00:34:29.200 | - Einstein would be proud.

00:34:30.440 | - Einstein, he might be appalled.

00:34:32.060 | I don't know.

00:34:32.900 | He didn't like quantum mechanics very much,

00:34:34.520 | but he would certainly be struck by it.

00:34:36.520 | I think that may be, at this time,

00:34:40.040 | its biggest contribution to physics in illustrating,

00:34:44.200 | almost definitively, that quantum mechanics and gravity

00:34:46.760 | are very closely related

00:34:48.600 | and not inconsistent with each other.

00:34:50.940 | - Is there a possibility of something deeper,

00:34:53.800 | more profound, that still is consistent with string theory,

00:34:58.800 | but is deeper, that is to be found?

00:35:03.080 | - Well, you could ask the same thing about quantum mechanics.

00:35:04.880 | Is there something-- - Exactly.

00:35:05.720 | - Yeah, yeah.

00:35:06.960 | I think string theory is just an example

00:35:09.040 | of a quantum mechanical system

00:35:11.000 | that contains both gravitation and quantum mechanics.

00:35:15.660 | So is there something underlying quantum mechanics?

00:35:19.760 | - Perhaps something deterministic.

00:35:21.560 | - Perhaps something deterministic.

00:35:23.920 | My friend, Gerard Etoft, whose name you may know,

00:35:27.360 | he's a very famous physicist, Dutch,

00:35:30.400 | not as famous as he should be, but--

00:35:32.200 | - Hard to spell his name.

00:35:35.160 | - It's hard to say his name.

00:35:36.280 | No, it's easy to spell his name.

00:35:37.480 | An apostrophe, he's the only person I know

00:35:39.280 | whose name begins with an apostrophe.

00:35:41.120 | And he's one of my heroes in physics.

00:35:44.320 | He's a little younger than me,

00:35:45.200 | but he's nevertheless one of my heroes.

00:35:47.640 | Etoft believes that there is some substructure

00:35:52.640 | to the world which is classical in character,

00:35:58.880 | deterministic in character,

00:36:01.120 | which somehow, by some mechanism

00:36:03.600 | that he has a hard time spelling out,

00:36:06.920 | emerges as quantum mechanics.

00:36:09.020 | I don't.

00:36:11.520 | - The wave function is somehow emergent.

00:36:13.800 | - The wave function, not just the wave function,

00:36:16.200 | but the whole mechan, the whole thing

00:36:18.720 | that goes with quantum mechanics,

00:36:19.960 | uncertainty, entanglement, all these things are emergent.

00:36:24.000 | - So you think quantum mechanics is the bottom of the well?

00:36:26.960 | - Here I think is where you have to be humble.

00:36:33.160 | Here's where humility comes.

00:36:34.440 | I don't think anybody should say

00:36:35.720 | anything is the bottom of the well at this time.

00:36:38.820 | I think we can reasonably say,

00:36:42.300 | I can reasonably say when I look into the well,

00:36:47.240 | I can't see past quantum mechanics.

00:36:50.500 | I don't see any reason for there to be anything

00:36:52.840 | beyond quantum mechanics.

00:36:55.080 | I think Etoft has asked very interesting

00:36:58.240 | and deep questions I don't like his answers.

00:37:00.840 | - Well, again, let me ask,

00:37:03.760 | if we look at the deepest nature of reality,

00:37:06.560 | whether it's deterministic

00:37:08.320 | or when observed as probabilistic,

00:37:13.080 | what does that mean for our human level

00:37:16.980 | of ideas of free will?

00:37:18.340 | Is there any connection whatsoever from this perception,

00:37:23.080 | perhaps illusion of free will that we have

00:37:25.800 | and the fundamental nature of reality?

00:37:27.760 | - The only thing I can say is I am puzzled by that

00:37:31.400 | as much as you are.

00:37:32.680 | - The illusion of it.

00:37:33.520 | - The illusion of consciousness,

00:37:36.120 | the illusion of free will, the illusion of self.

00:37:40.020 | - Does that connect to--

00:37:43.400 | - How can a physical system do that?

00:37:45.520 | And I am as puzzled as anybody.

00:37:48.880 | - There's echoes of it in the observer effect.

00:37:51.760 | - Yeah.

00:37:52.600 | - So do you understand what it means to be an observer?

00:37:55.360 | - I understand it at a technical level.

00:37:57.720 | An observer is a system with enough degrees of freedom

00:38:00.520 | that it can record information

00:38:02.320 | and which can become entangled

00:38:04.000 | with the thing that it's measuring.

00:38:05.760 | Entanglement is the key.

00:38:07.320 | When a system which we call an apparatus or an observer,

00:38:12.040 | same thing, interacts with the system that it's observing,

00:38:16.340 | it doesn't just look at it,

00:38:19.160 | it becomes physically entangled with it.

00:38:21.600 | And it's that entanglement which we call an observation

00:38:24.480 | or a measurement.

00:38:26.520 | Now does that satisfy me personally as an observer?

00:38:30.440 | - Mm-hmm.

00:38:31.280 | - Yes and no.

00:38:33.380 | I find it very satisfying that we have

00:38:34.860 | a mathematical representation of what it means

00:38:38.280 | to observe a system.

00:38:40.360 | - You are observing stuff right now.

00:38:42.360 | - Yeah.

00:38:43.200 | - The conscious level.

00:38:44.020 | - Right.

00:38:44.860 | - Do you think there's echoes of that kind of entanglement

00:38:48.120 | in our macro scale?

00:38:49.480 | - Yes, absolutely, for sure.

00:38:52.240 | We're entangled with, quantum mechanically entangled

00:38:55.040 | with everything in this room.

00:38:56.780 | If we weren't, then we would just,

00:38:59.900 | well, we wouldn't be observing it.

00:39:01.780 | But on the other hand, you can ask,

00:39:05.660 | do I really, am I really comfortable with it?

00:39:10.300 | And I'm uncomfortable with it in the same way

00:39:12.580 | that I can never get comfortable with five dimensions.

00:39:15.340 | My brain isn't wired for it.

00:39:17.540 | - Are you comfortable with four dimensions?

00:39:21.420 | - A little bit more because I can always imagine

00:39:24.260 | the fourth dimension is time.

00:39:26.380 | - So the arrow of time, are you comfortable with that arrow?

00:39:30.020 | Do you think time is an emergent phenomena

00:39:32.020 | or is it fundamental to nature?

00:39:33.740 | - That is a big question in physics right now.

00:39:37.620 | All the physics that we do, or at least at the people

00:39:41.780 | that I am comfortable with talking to, my friends.

00:39:44.780 | - Yeah.

00:39:48.180 | - My friends.

00:39:49.460 | We all ask the same question that you just asked.

00:39:51.940 | Space, we have a pretty good idea, is emergent.

00:39:55.300 | And it emerges out of entanglement and other things.

00:40:00.300 | Time always seems to be built into our equations

00:40:03.940 | as just what Newton pretty much would have thought.

00:40:06.700 | Newton modified a little bit by Einstein

00:40:09.140 | would have called time.

00:40:10.380 | And mostly in our equations, it is not emergent.

00:40:16.980 | Time in physics is completely symmetric,

00:40:21.580 | forward and backward. - Right, it's symmetric.

00:40:23.420 | So you don't really need to think about the arrow of time

00:40:27.300 | for most physical phenomena.

00:40:29.240 | - Most microscopic phenomena, no.

00:40:33.380 | It's only when the phenomena involve systems

00:40:35.620 | which are big enough for thermodynamics to become important,

00:40:38.940 | for entropy to become important.

00:40:40.840 | For a small system, entropy is not a good concept.

00:40:46.020 | And entropy is something which emerges out of large numbers.

00:40:52.840 | It's a probabilistic idea, it's a statistical idea,

00:40:56.600 | and it's a thermodynamic idea.

00:40:58.500 | Thermodynamics requires lots and lots

00:41:00.860 | and lots of little substructures.

00:41:02.680 | So it's not until you emerge at the thermodynamic level

00:41:09.140 | that there's an arrow of time.

00:41:11.720 | Do we understand it?

00:41:13.540 | Yeah, I think we understand better

00:41:15.820 | than most people think they have.

00:41:17.180 | Most people say they think we understand it.

00:41:19.340 | Yeah, I think we understand it.

00:41:21.300 | It's a statistical idea.

00:41:23.940 | - You mean like second law of thermodynamics,

00:41:26.420 | entropy and so on? - Yeah, yeah.

00:41:27.260 | Take a pack of cards and you fling it in the air

00:41:29.740 | and you look what happens to it.

00:41:31.460 | It gets random. - Yeah, but we understand it.

00:41:33.380 | - It doesn't go from random to simple.

00:41:36.220 | It goes from simple to random.

00:41:37.740 | - But do you think it ever breaks down?

00:41:41.820 | - What I think you can do is in a laboratory setting,

00:41:46.180 | you can take a system which is somewhere intermediate

00:41:49.020 | between being small and being large

00:41:51.180 | and make it go backward.

00:41:56.100 | A thing which looks like it only wants to go forward

00:41:59.580 | because of statistical mechanical reasons,

00:42:01.660 | because of the second law,

00:42:03.900 | you can very, very carefully manipulate it

00:42:07.120 | to make it run backward.

00:42:09.100 | I don't think you can take an egg,

00:42:10.660 | a Humpty Dumpty who fell on the floor and reverse that.

00:42:15.000 | But you can, in a very controlled situation,

00:42:18.500 | you can take systems which appear to be evolving

00:42:22.780 | statistically toward randomness,

00:42:25.380 | stop them, reverse them and make them go back.

00:42:29.280 | - What's the intuition behind that?

00:42:30.780 | How do we do that?

00:42:31.940 | How do we reverse it?

00:42:33.500 | You're saying a closed system?

00:42:35.580 | - Yeah, pretty much closed system, yes.

00:42:38.340 | - Did you just say that time travel is possible?

00:42:41.580 | - No, I didn't say time travel is possible.

00:42:44.060 | I said you can make a system go backward.

00:42:45.960 | - In time.

00:42:46.800 | - You can make it go back, you can make it reverse

00:42:49.020 | its steps, you can make it reverse its trajectory.

00:42:51.560 | - Yeah.

00:42:52.440 | How do we do it?

00:42:53.280 | What's the intuition there?

00:42:54.660 | Does it have, is it just a fluke thing

00:42:58.780 | that we can do at a small scale in the lab

00:43:00.880 | that doesn't have--

00:43:01.720 | - What I'm saying is you can do it

00:43:02.680 | on a little bit better than a small scale.

00:43:05.300 | You can certainly do it with a simple small system.

00:43:10.300 | Small systems don't have any sense of the arrow of time.

00:43:14.300 | Atoms, atoms are no sense of an arrow of time.

00:43:19.300 | They're completely reversible.

00:43:22.420 | It's only when you have, you know,

00:43:24.480 | the second law of thermodynamics is the law of large numbers.

00:43:28.520 | - So you can break the law because it's not--

00:43:30.840 | - You can break it, but it's hard.

00:43:33.920 | It requires great care.

00:43:36.120 | The bigger the system is, the more care,

00:43:38.480 | the more, the harder it is.

00:43:40.800 | You have to overcome what's called chaos.

00:43:43.960 | And that's hard.

00:43:45.640 | And it requires more and more precision.

00:43:47.680 | For 10 particles, you might be able to do it

00:43:50.120 | with some effort.

00:43:54.040 | For 100 particles, it's really hard.

00:43:56.820 | For 1,000 or a million particles, forget it.

00:43:59.360 | But not for any fundamental reason,

00:44:01.260 | just because it's technologically too hard

00:44:03.760 | to make the system go backward.

00:44:05.700 | - So, no time travel for engineering reasons.

00:44:13.880 | - No, no, no, no.

00:44:15.320 | What is time travel?

00:44:16.920 | Time travel to the future, that's easy.

00:44:20.440 | You just close your eyes, go to sleep,

00:44:22.440 | and you wake up in the future.

00:44:23.520 | - Yeah, yeah.

00:44:24.640 | A good nap gets you there, yeah.

00:44:25.960 | - A good nap gets you there, right.

00:44:27.400 | - But in reversing the second law of thermodynamics,

00:44:32.280 | the going backward in time,

00:44:33.840 | for anything that's human scale

00:44:36.780 | is a very difficult engineering effort.

00:44:40.020 | - I wouldn't call that time travel

00:44:41.440 | because it gets too mixed up

00:44:43.040 | with what science fiction calls time travel.

00:44:46.040 | This is just the ability to reverse a system.

00:44:51.040 | You take the system, and you reverse the direction

00:44:55.680 | of motion of every molecule in it.

00:44:58.000 | You can do it with one molecule.

00:45:00.940 | If you find a particle moving in a certain direction,

00:45:03.500 | let's not say a particle, a baseball,

00:45:05.720 | you stop it dead, and then you simply reverse its motion,

00:45:10.960 | in principle that's not too hard,

00:45:12.840 | and it'll go back along its trajectory

00:45:15.240 | in the backward direction.

00:45:16.440 | - Just running the program backwards.

00:45:18.200 | - Running the program backward.

00:45:20.440 | If you have two baseballs colliding,

00:45:22.400 | well you can do it, but you have to be very, very careful

00:45:25.820 | to get it just right.

00:45:26.920 | If you have 10 baseballs, really, really, or better yet,

00:45:32.400 | 10 billiard balls on an idealized,

00:45:36.200 | frictionless billiard table.

00:45:38.840 | Okay, so you start the balls all in a triangle, right?

00:45:41.840 | And you whack 'em.

00:45:43.480 | Depending on the game you're playing,

00:45:44.640 | you either whack 'em or you're really careful,

00:45:46.040 | but you whack 'em, and they go flying off

00:45:49.360 | in all possible directions.

00:45:50.880 | Try to reverse that.

00:45:53.200 | Try to reverse that.

00:45:55.720 | Imagine trying to take every billiard ball,

00:45:57.720 | stopping it dead at some point,

00:46:00.360 | and reversing its motion so that it was going

00:46:02.420 | in the opposite direction.

00:46:04.240 | If you did that with tremendous care,

00:46:07.400 | it would reassemble itself back into the triangle.

00:46:11.560 | Okay, that is a fact, and you can probably do it

00:46:15.780 | with two billiard balls, maybe with three billiard balls

00:46:18.320 | if you're really lucky.

00:46:19.960 | But what happens is as the system gets more and more

00:46:22.440 | complicated, you have to be more and more precise

00:46:26.160 | not to make the tiniest error,

00:46:27.920 | because the tiniest errors will get magnified,

00:46:31.000 | and you'll simply not be able to do the reversal.

00:46:34.960 | So yeah, but I wouldn't call that time travel.

00:46:38.600 | - Yeah, that's something else.

00:46:39.720 | But if you think of it, it just made me think,

00:46:43.280 | if you think the unrolling of state that's happening

00:46:48.200 | as a program, if we look at the world,

00:46:52.880 | so the idea of looking at the world as a simulation,

00:46:56.480 | as a computer, but it's not a computer,

00:47:00.060 | it's just a single program.

00:47:03.240 | A question arises that might be useful.

00:47:05.800 | How hard is it to have a computer that runs the universe?

00:47:10.800 | - Okay, so there are mathematical universes

00:47:16.560 | that we know about, one of them is called

00:47:20.760 | anti-de Sitter space, where we,

00:47:24.640 | and it's quantum mechanics, where I think we could

00:47:30.000 | simulate it in a computer, in a quantum computer.

00:47:34.120 | Classical computer, all you can do is solve its equations,

00:47:36.600 | you can't make it work like the real system.

00:47:39.120 | If we could build a quantum computer, a big enough one,

00:47:41.560 | a robust enough one, we could probably simulate

00:47:46.360 | a universe, a small version of an anti-de Sitter universe.

00:47:52.280 | Anti-de Sitter is a kind of a cosmology.

00:47:55.760 | All right, so I think we know how to do that.

00:48:00.080 | The trouble is, the universe that we live in

00:48:02.320 | is not the anti-de Sitter geometry,

00:48:04.800 | it's the de Sitter geometry, and we don't really understand

00:48:09.080 | its quantum mechanics at all.

00:48:11.080 | So at the present time, I would say we wouldn't have

00:48:12.960 | the vaguest idea how to simulate a universe

00:48:15.400 | similar to our own.

00:48:16.480 | No, we can ask, could we build in the laboratory

00:48:21.080 | a small version, a quantum mechanical version,

00:48:27.520 | a collection of quantum computers entangled

00:48:30.480 | and coupled together, which would reproduce

00:48:34.440 | the phenomena that go on in the universe,

00:48:38.160 | even on a small scale?

00:48:40.440 | Yes, if it were anti-de Sitter space.

00:48:43.080 | No, if it's de Sitter space.

00:48:44.720 | - Can you slightly describe de Sitter space

00:48:47.440 | and anti-de Sitter space?

00:48:48.680 | - Yeah.

00:48:49.880 | - What are the geometric properties of--

00:48:51.440 | - They differ by the sign of a single constant

00:48:54.880 | called the cosmological constant.

00:48:57.520 | One of them is negatively curved,

00:49:01.760 | the other is positively curved.

00:49:03.680 | Anti-de Sitter space, which is the negatively curved one,

00:49:08.240 | you can think of as an isolated system

00:49:11.360 | in a box with reflecting walls.

00:49:14.600 | You could think of it as a system of,

00:49:16.480 | a quantum mechanical system isolated

00:49:19.240 | in an isolated environment.

00:49:21.800 | De Sitter space is the one we really live in,

00:49:23.720 | and that's the one that's exponentially expanding.

00:49:26.880 | Exponential expansion, dark energy,

00:49:30.040 | whatever you want to call it,

00:49:31.520 | and we don't understand that mathematically.

00:49:34.300 | - Do we understand--

00:49:36.960 | - Not everybody would agree with me,

00:49:38.200 | but I don't understand.

00:49:39.560 | They would agree with me, they definitely would agree

00:49:41.640 | with me that I don't understand it.

00:49:43.440 | (laughing)

00:49:44.740 | - What about, is there an understanding

00:49:47.000 | of the birth, the origin, the Bing Bang?

00:49:50.320 | - No, no, no, there's theories.

00:49:53.160 | There are theories.

00:49:55.840 | My favorite is the one called eternal inflation.

00:49:59.000 | - The infinity can be on both sides,

00:50:00.680 | on one of the sides, and none of the sides.

00:50:02.600 | So what's eternal infinity?

00:50:05.520 | - Okay.

00:50:06.360 | Infinity on both sides.

00:50:11.120 | - Oh boy.

00:50:13.920 | - Yeah, yeah, that's--

00:50:15.480 | - Why is that your favorite?

00:50:16.520 | 'Cause it's the most, just mind-blowing?

00:50:21.520 | - No.

00:50:22.380 | - 'Cause we want a beginning.

00:50:23.200 | - No, why do we want a beginning?

00:50:25.960 | I--

00:50:26.800 | - In practice, there was a beginning, of course.

00:50:28.400 | In practice, there was a beginning.

00:50:31.280 | But could it have been a random fluctuation

00:50:36.280 | in an otherwise infinite time?

00:50:39.360 | Maybe.

00:50:40.200 | In any case, the eternal inflation theory,

00:50:45.880 | I think if correctly understood,

00:50:47.320 | would be infinite in both directions.

00:50:49.280 | - How do you think about infinity?

00:50:52.880 | - Oh God.

00:50:54.000 | - So, okay, of course you can think about it mathematically.

00:50:57.560 | - I just finished this discussion

00:50:59.640 | with my friend Sergey Brin.

00:51:01.200 | - Yes.

00:51:02.020 | - How do you think about infinity?

00:51:02.860 | I say, well, Sergey Brin is infinitely rich.

00:51:05.360 | (laughing)

00:51:08.000 | - How do you test that hypothesis?

00:51:09.400 | Okay.

00:51:10.240 | That's such a good line.

00:51:13.320 | Right.

00:51:14.160 | Yeah, so there's really no way

00:51:17.200 | to visualize some of these things.

00:51:18.880 | - Yeah, no, this is a very good question.

00:51:22.400 | Does physics have any,

00:51:24.680 | does infinity have any place in physics?

00:51:27.360 | - Right.

00:51:28.200 | - Right, and all I can say is, very good question.

00:51:32.600 | - So what do you think of the recent first image

00:51:39.400 | of a black hole visualized from the Event Horizon Telescope?

00:51:43.040 | - It's an incredible triumph of science.

00:51:45.680 | In itself, the fact that there are black holes

00:51:47.760 | which collide is not a surprise.

00:51:50.640 | And they seem to work exactly

00:51:52.680 | the way they're supposed to work.

00:51:54.800 | Will we learn a great deal from it?

00:51:57.080 | I don't know, I can't, we might.

00:52:00.180 | But the kind of things we'll learn

00:52:01.320 | won't really be about black holes.

00:52:03.320 | Why there are black holes in nature

00:52:09.520 | of that particular mass scale and why they're so common

00:52:12.800 | may tell us something about the structure,

00:52:15.720 | evolution of structure in the universe.

00:52:18.480 | But I don't think it's gonna tell us

00:52:19.560 | anything new about black holes.

00:52:22.040 | But it's a triumph in the sense that you go back 100 years

00:52:25.480 | and it was a continuous development,

00:52:28.240 | general relativity, the discovery of black holes,

00:52:31.520 | LIGO, the incredible technology that went into LIGO.

00:52:35.540 | It is something that I never would have believed

00:52:42.640 | was gonna happen 30, 40 years ago.

00:52:47.800 | And I think it's a magnificent structure,

00:52:51.840 | a magnificent thing, this evolution of general relativity,

00:52:56.840 | LIGO, high precision, ability to measure things

00:53:04.000 | on a scale of 10 to the minus 21.

00:53:06.300 | So, yeah.

00:53:09.440 | - So you're just in awe-- - Astonishing.

00:53:11.320 | Just in awe. - That this path

00:53:12.800 | took us to this picture.

00:53:14.840 | Is it different?

00:53:15.880 | You've thought a lot about black holes.

00:53:19.920 | How did you visualize them in your mind?

00:53:22.460 | And is the picture different than you realized it?

00:53:26.320 | - No, it simply confirmed.

00:53:28.780 | It's a magnificent triumph to have confirmed--

00:53:32.280 | - Confirmed. - A direct observation

00:53:34.960 | that Einstein's theory of gravity

00:53:37.400 | at the level of black hole collisions actually works.

00:53:42.440 | It's awesome, it is really awesome.

00:53:45.480 | You know, I know some of the people

00:53:46.600 | who are involved in that.

00:53:48.240 | They're just ordinary people.

00:53:50.000 | And the idea that they could carry this out,

00:53:53.680 | I just, I'm shocked.

00:53:56.500 | - Yeah, just these little homo sapiens.

00:53:59.400 | - Yeah, just these little monkeys.

00:54:01.320 | - Got together and took a picture of--

00:54:04.840 | - Slightly advanced limers, I think.

00:54:06.740 | - What kind of questions can science not currently answer

00:54:11.360 | but you hope might be able to soon?

00:54:13.480 | - Well, you've already addressed them.

00:54:15.120 | What is consciousness, for example?

00:54:17.040 | - You think that's within the reach of science?

00:54:19.440 | - I think it's somewhat within the reach of science,

00:54:21.680 | but I think now it's in the hands

00:54:23.720 | of the computer scientists and the neuroscientists.

00:54:27.160 | - Not a physicist.

00:54:28.520 | - Perhaps at some point. - With the help.

00:54:30.840 | - But I think physicists will try to simplify it down

00:54:34.120 | to something that they can use their methods

00:54:36.360 | and maybe they're not appropriate.

00:54:38.600 | Maybe we simply need to do more machine learning

00:54:43.600 | on bigger scales, evolve machines.

00:54:47.840 | Machines not only that learn,

00:54:49.400 | but evolve their own architecture

00:54:51.360 | as a process of learning, evolving architecture,

00:54:54.360 | not under our control, only partially under our control,

00:54:57.000 | but under the control of a machine learning.

00:54:59.260 | I'll tell you another thing that I find awesome.

00:55:02.760 | You know this Google thing

00:55:04.840 | that they taught computers how to play chess?

00:55:07.520 | - Yeah, yeah. - Okay.

00:55:08.880 | They taught computers how to play chess,

00:55:10.660 | not by teaching them how to play chess,

00:55:12.440 | but just having them play against each other.

00:55:14.400 | - Against each other, self-play.

00:55:15.440 | - Against each other.

00:55:16.280 | This is a form of evolution.

00:55:17.840 | These machines evolved.

00:55:21.000 | They evolved in intelligence.

00:55:22.940 | They evolved in intelligence

00:55:27.400 | without anybody telling them how to do it.

00:55:30.880 | They were not engineered.

00:55:32.560 | They just played against each other

00:55:34.000 | and got better and better and better.

00:55:36.680 | That makes me think that machines can evolve intelligence.

00:55:41.680 | What exact kind of intelligence, I don't know,

00:55:46.600 | but in understanding that better and better,

00:55:49.040 | maybe we'll get better clues

00:55:50.720 | as to what goes on in our own intelligence.

00:55:53.040 | - Well, life and intelligence is, last question,

00:55:55.920 | what kind of questions can science not currently answer

00:55:59.880 | and may never be able to answer?

00:56:01.880 | - Yeah.

00:56:05.660 | Is there an intelligence out there

00:56:07.340 | that's underlies the whole thing?

00:56:09.280 | You can call them with a G word if you want.

00:56:12.040 | I can say, are we a computer simulation with a purpose?

00:56:16.960 | Is there an agent, an intelligent agent

00:56:22.600 | that underlies or is responsible for the whole thing?

00:56:27.200 | Does that intelligent agent satisfy the laws of physics?

00:56:30.600 | Does it satisfy the laws of quantum mechanics?

00:56:32.640 | Is it made of atoms and molecules?

00:56:34.480 | Yeah, there's a lot of questions.

00:56:36.560 | And I don't see, it seems to me a real question.

00:56:41.080 | - It's an answerable question.

00:56:43.520 | - Well, I don't know if it's answerable.

00:56:44.800 | The questions have to be answerable to be real.

00:56:47.360 | Some philosophers would say that a question

00:56:52.240 | is not a question unless it's answerable.

00:56:54.520 | This question doesn't seem to me answerable

00:56:58.320 | by any known method, but it seems to me real.

00:57:04.040 | - There's no better place to end.

00:57:07.200 | Leonard, thank you so much for talking today.

00:57:08.760 | - Okay, good.

00:57:09.600 | (audience applauding)

00:57:12.760 | (upbeat music)

00:57:15.340 | (upbeat music)

00:57:17.920 | (upbeat music)

00:57:20.500 | (upbeat music)

00:57:23.080 | (upbeat music)

00:57:25.660 | [BLANK_AUDIO]

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

Chapters