Richard Feynman on Computation (Stephen Wolfram)

00:00:00.000 | - When you were at Caltech, did you get to interact

00:00:05.000 | with Richard Feynman at all?

00:00:07.240 | Do you have any memories of Richard?

00:00:08.880 | - We worked together quite a bit actually.

00:00:11.920 | In fact, both when I was at Caltech

00:00:14.960 | and after I left Caltech, we were both consultants

00:00:18.540 | at this company called Thinking Machines Corporation,

00:00:20.600 | which was just down the street from here actually.

00:00:23.360 | It was ultimately ill-fated company,

00:00:25.140 | but I used to say this company is not gonna work

00:00:28.620 | with the strategy they have and Dick Feynman always used

00:00:30.760 | to say, "What do we know about running companies?

00:00:32.920 | "Just let them run their company."

00:00:34.960 | But anyway, he was not into that kind of thing

00:00:39.960 | and he always thought that my interest in doing things

00:00:43.020 | like running companies was a distraction so to speak.

00:00:47.080 | And for me, it's a mechanism to have a more effective

00:00:52.080 | machine for actually figuring things out

00:00:56.660 | and getting things to happen.

00:00:57.880 | - Did he think of it 'cause essentially what you used,

00:01:01.080 | you did with the company, I don't know if you were thinking

00:01:02.840 | of it that way, but you're creating tools

00:01:06.120 | to empower the exploration of the university.

00:01:11.120 | Do you think, did he--

00:01:13.480 | - Did he understand that point?

00:01:14.960 | - The point of tools of--

00:01:17.140 | - I think not as well as he might have done.

00:01:19.040 | I mean, I think that, but he was actually my first company,

00:01:23.620 | which was also involved with, well, was involved

00:01:27.000 | with more mathematical computation kinds of things.

00:01:29.700 | He was quite, he had lots of advice

00:01:34.240 | about the technical side of what we should do and so on.

00:01:37.780 | - Do you have examples, memories, or thoughts that--

00:01:39.800 | - Oh yeah, yeah, he had all kinds of,

00:01:41.560 | look, in the business of doing sort of,

00:01:45.360 | one of the hard things in math is doing integrals

00:01:47.200 | and so on, right?

00:01:48.240 | And so he had his own elaborate ways to do integrals

00:01:51.560 | and so on, he had his own ways of thinking about,

00:01:53.300 | sort of getting intuition about how math works.

00:01:56.400 | And so his sort of meta idea was,

00:02:00.160 | take those intuitional methods and make a computer

00:02:02.760 | follow those intuitional methods.

00:02:04.720 | Now it turns out, for the most part,

00:02:07.580 | like when we do integrals and things,

00:02:09.520 | what we do is we build this kind of bizarre industrial

00:02:12.840 | machine that turns every integral into, you know,

00:02:15.560 | products of major G functions and generates

00:02:17.960 | this very elaborate thing.

00:02:19.760 | And actually the big problem is turning the results

00:02:22.220 | into something a human will understand.

00:02:23.800 | It's not, quote, doing the integral.

00:02:26.080 | And actually, Feynman did understand that to some extent.

00:02:28.640 | And I'm embarrassed to say, he once gave me this big pile

00:02:32.520 | of, you know, calculational methods for particle physics

00:02:35.560 | that he worked out in the '50s, and he said,

00:02:37.160 | you know, it's more used to you than to me type thing.

00:02:39.000 | And I was like, I've intended to look at it

00:02:41.600 | and give it back, and it's still in my files now.

00:02:43.760 | So it's, but that's what happens when it's finiteness

00:02:48.160 | of human lives.

00:02:49.120 | It, you know, maybe if he'd lived another 20 years,

00:02:51.880 | I would have remembered to give it back.

00:02:53.660 | But I think it's, you know, that was his attempt

00:02:57.960 | to systematize the ways that one does integrals

00:03:02.560 | that show up in particle physics and so on.

00:03:04.340 | Turns out the way we've actually done it

00:03:06.480 | is very different from that way.

00:03:08.120 | - What do you make of that difference between,

00:03:09.980 | so Feynman was actually quite remarkable at creating

00:03:13.600 | sort of intuitive, like diving in, you know,

00:03:17.040 | creating intuitive frameworks for understanding

00:03:19.400 | difficult concepts.

00:03:20.840 | Is--

00:03:21.680 | - I'm smiling because, you know, the funny thing

00:03:24.260 | about him was that the thing he was really, really,

00:03:27.300 | really good at is calculating stuff.

00:03:29.820 | And, but he thought that was easy

00:03:31.540 | because he was really good at it.

00:03:33.860 | And so he would do these things where he would calculate

00:03:36.580 | some, do some complicated calculation

00:03:40.060 | in quantum field theory, for example,

00:03:41.700 | come out with a result.

00:03:43.320 | Wouldn't tell anybody about the complicated calculation

00:03:45.380 | 'cause he thought that was easy.

00:03:46.620 | He thought the really impressive thing

00:03:48.440 | was to have the simple intuition

00:03:50.020 | about how everything works.

00:03:52.080 | So he invented that at the end.

00:03:54.360 | And, you know, because he'd done this calculation

00:03:56.440 | and knew how it worked, it was a lot easier.

00:03:59.240 | It's a lot easier to have good intuition

00:04:00.920 | when you know what the answer is.

00:04:02.720 | And then he would just not tell anybody

00:04:05.000 | about these calculations.

00:04:06.000 | And he wasn't meaning that maliciously, so to speak.

00:04:08.680 | It's just, he thought that was easy.

00:04:11.320 | And that's, you know, that led to areas

00:04:13.820 | where people were just completely mystified

00:04:15.600 | and they kind of followed his intuition,

00:04:17.440 | but nobody could tell why it worked

00:04:19.520 | because actually the reason it worked

00:04:21.160 | was 'cause he'd done all these calculations

00:04:22.580 | and he knew that it would work.

00:04:24.460 | And, you know, when I, he and I worked a bit

00:04:26.700 | on quantum computers actually back in 1980, '81,

00:04:31.200 | but before anybody had heard of those things.

00:04:33.580 | And, you know, the typical mode of,

00:04:35.640 | I mean, he always used to say,

00:04:38.140 | and I now think about this 'cause I'm about the age

00:04:40.100 | that he was when I worked with him.

00:04:42.340 | And, you know, I see that people who are one third my age,

00:04:44.900 | so to speak, and he was always complaining

00:04:47.420 | that I was one third his age.

00:04:48.940 | (both laughing)

00:04:50.520 | Various things, but, you know,

00:04:52.940 | he would do some calculation by hand, you know,

00:04:56.260 | blackboard and things, come up with some answer.

00:04:58.960 | I'd say, "I don't understand this."

00:05:01.320 | You know, I do something with a computer

00:05:03.480 | and he'd say, you know, "I don't understand this."

00:05:07.040 | So there'd be some big argument about what was,

00:05:09.540 | you know, what was going on, but it was always,

00:05:11.940 | and I think actually many of the things

00:05:16.340 | that we sort of realized about quantum computing

00:05:19.940 | that were sort of issues that have to do particularly

00:05:21.780 | with the measurement process are kind of still issues today.

00:05:25.540 | And I kind of find it interesting.

00:05:26.980 | It's a funny thing in science that these, you know,

00:05:30.140 | that there's a remarkable, it happens in technology too,

00:05:33.620 | there's a remarkable sort of repetition of history

00:05:36.860 | that ends up occurring.

00:05:38.580 | Eventually things really get nailed down,

00:05:40.760 | but it often takes a while

00:05:42.500 | and it often things come back decades later.

00:05:46.020 | Well, for example, I could tell a story,

00:05:48.260 | actually happened right down the street from here.

00:05:50.760 | When we were both at Thinking Machines,

00:05:53.220 | I had been working on this particular cellular automaton

00:05:56.820 | called Rule 30 that has this feature that it,

00:05:59.400 | from very simple initial conditions,

00:06:01.700 | it makes really complicated behavior, okay?

00:06:04.440 | So, and actually of all silly physical things,

00:06:08.780 | using this big parallel computer

00:06:11.780 | called the Connection Machine that that company was making,

00:06:15.380 | I generated this giant printout of Rule 30 on very,

00:06:19.060 | on actually on the same kind of printer

00:06:21.760 | that people use to make layouts for microprocessors.

00:06:26.760 | So one of these big, you know,

00:06:29.460 | large format printers with high resolution and so on.

00:06:33.000 | So, okay, so we print this out, lots of very tiny cells.

00:06:36.940 | And so there was sort of a question of how,

00:06:39.740 | some features of that pattern.

00:06:42.620 | And so it was very much a physical, you know,

00:06:45.500 | on the floor with meter rules

00:06:46.860 | trying to measure different things.

00:06:48.820 | So, so Feynman kind of takes me aside.

00:06:51.860 | We've been doing that for a little while

00:06:53.060 | and takes me aside and he says,

00:06:54.820 | "I just wanna know this one thing."

00:06:56.300 | He says, "I wanna know, how did you know

00:06:58.580 | that this Rule 30 thing would produce

00:07:01.340 | all this really complicated behavior

00:07:02.660 | that is so complicated that we're, you know,

00:07:05.220 | going around with this big printout and so on?"

00:07:07.580 | And I said, "Well, I didn't know.

00:07:09.900 | I just enumerated all the possible rules

00:07:12.120 | and then observed that that's what happened."

00:07:14.860 | He said, "Oh, I feel a lot better."

00:07:17.220 | You know, I thought you had some intuition

00:07:19.160 | that he didn't have that would let one.

00:07:21.780 | I said, "No, no, no, no intuition,

00:07:23.420 | just experimental science."

00:07:25.300 | - Oh, that's such a beautiful sort of dichotomy there

00:07:29.420 | of that's exactly what you showed

00:07:31.060 | is you really can't have an intuition

00:07:33.260 | about an irreducible, I mean, you have to run it.

00:07:36.020 | - Yes, that's right.

00:07:36.860 | - That's so hard for us humans

00:07:38.180 | and especially brilliant physicists

00:07:41.820 | like Feynman to say that you can't have

00:07:44.640 | a compressed, clean intuition

00:07:48.600 | about how the whole thing works.

00:07:50.600 | - Yes, yes.

00:07:51.900 | No, he was, I mean, I think he was sort of on the edge

00:07:54.640 | of understanding that point about computation.

00:07:56.960 | And I think he found that,

00:07:58.640 | I think he always found computation interesting.

00:08:01.320 | And I think that was sort of what he was

00:08:03.040 | a little bit poking at.

00:08:04.740 | I mean, that intuition, you know,

00:08:07.120 | the difficulty of discovering things like even you say,

00:08:09.640 | oh, you know, you just enumerate all the cases

00:08:11.520 | and you just find one that does something interesting,

00:08:13.220 | right, sounds very easy.

00:08:15.220 | Turns out like I missed it when I first saw it

00:08:18.300 | because I had kind of an intuition

00:08:20.300 | that said it shouldn't be there.

00:08:21.780 | And so I had kind of arguments,

00:08:23.100 | oh, I'm gonna ignore that case because whatever.

00:08:25.800 | And--

00:08:27.540 | - How did you have an open mind enough?

00:08:29.540 | Because you're essentially the same person

00:08:31.260 | as Richard Feynman, like the same kind of physics

00:08:33.500 | type of thinking.

00:08:34.700 | How did you find yourself having a sufficiently open mind

00:08:38.900 | to be open to watching rules and them revealing complexity?

00:08:43.280 | - Yeah, I think that's an interesting question.

00:08:44.620 | I've wondered about that myself

00:08:45.940 | 'cause it's kind of like, you know,

00:08:47.400 | you live through these things and then you say,

00:08:50.000 | what was the historical story?

00:08:51.600 | And sometimes the historical story that you realize

00:08:53.620 | after the fact was not what you lived through, so to speak.

00:08:57.040 | And so, you know, what I realized is I think what happened

00:09:01.920 | is, you know, I did physics kind of like

00:09:05.760 | reductionistic physics where you're throwing the universe

00:09:08.640 | and you're told go figure out what's going on inside it.

00:09:11.600 | And then I started building computer tools

00:09:14.600 | and I started building my first computer language,

00:09:17.000 | for example.

00:09:18.160 | And computer language is not like,

00:09:19.760 | it's sort of like physics in the sense that you have to take

00:09:22.320 | all those computations people want to do

00:09:24.280 | and kind of drill down and find the primitives

00:09:26.860 | that they can all be made of.

00:09:28.640 | But then you do something that's really different

00:09:30.320 | because you're just saying, okay, these are the primitives.

00:09:33.700 | Now, you know, hopefully they'll be useful to people.

00:09:36.280 | Let's build up from there.

00:09:37.700 | So you're essentially building an artificial universe

00:09:40.660 | in a sense where you make this language,

00:09:42.960 | you've got these primitives,

00:09:44.320 | you're just building whatever you feel like building.

00:09:47.240 | And that's, and so it was sort of interesting for me

00:09:50.080 | because from doing science where you're just throwing

00:09:52.360 | the universe as the universe is to then just being told,

00:09:56.760 | you know, you can make up any universe you want.

00:09:59.480 | And so I think that experience of making a computer language,

00:10:03.100 | which is essentially building your own universe,

00:10:05.040 | so to speak, is, you know, that's kind of the,

00:10:09.520 | that's what gave me a somewhat different attitude

00:10:12.560 | towards what might be possible.

00:10:13.960 | It's like, let's just explore what can be done

00:10:16.200 | in these artificial universes,

00:10:18.160 | rather than thinking the natural science way

00:10:21.240 | of let's be constrained by how the universe actually is.

00:10:23.800 | - Yeah, by being able to program, essentially you've,

00:10:26.880 | as opposed to being limited to just your mind and a pen,

00:10:31.000 | you now have, you've basically built another brain

00:10:34.360 | that you can use to explore the universe by,

00:10:36.400 | the computer program, you know, is a kind of a brain.

00:10:39.920 | - Right, and it's, well, it's, or a telescope,

00:10:42.120 | or, you know, it's a tool.

00:10:43.280 | It lets you see stuff.

00:10:45.040 | - But there's something fundamentally different

00:10:46.400 | between a computer and a telescope.

00:10:47.880 | I mean, it just, I'm hoping not to romanticize the notion,

00:10:52.880 | but it's more general, the computer is more general

00:10:55.760 | than the telescope. - It is, it is much more general.

00:10:56.600 | And it's, I think, I mean, this point about,

00:10:59.800 | you know, people say, oh, such and such a thing

00:11:04.280 | was almost discovered at such and such a time.

00:11:06.840 | The distance between, you know,

00:11:09.160 | the building the paradigm that allows you

00:11:10.760 | to actually understand stuff,

00:11:12.120 | or allows one to be open to seeing what's going on,

00:11:15.120 | that's really hard.

00:11:16.560 | And, you know, I think in, I've been fortunate in my life

00:11:20.760 | that I've spent a lot of my time

00:11:22.120 | building computational language,

00:11:24.280 | and that's an activity that in a sense works

00:11:28.640 | by sort of having to kind of create

00:11:33.400 | another level of abstraction and kind of be open

00:11:35.720 | to different kinds of structures.

00:11:37.520 | But, you know, it's always, I mean, I'm fully aware of,

00:11:41.860 | I suppose, the fact that I have seen it a bunch of times

00:11:45.360 | of how easy it is to miss the obvious, so to speak,

00:11:48.640 | that at least is factored into my attempt

00:11:51.640 | to not miss the obvious, although it may not succeed.

00:11:55.200 | (whooshing)

00:11:57.520 | (whooshing)

00:11:59.840 | (whooshing)

00:12:02.160 | (whooshing)

00:12:04.480 | (whooshing)

00:12:06.800 | (whooshing)

00:12:09.120 | (whooshing)

00:12:11.440 | [BLANK_AUDIO]

Richard Feynman on Computation (Stephen Wolfram) | AI Podcast Clips

Chapters