episode
stringlengths
45
100
text
stringlengths
1
528
timestamp_link
stringlengths
56
56
Guido van Rossum: Python | Lex Fridman Podcast #6
humans around them are going to do, or at least they have a model of what those humans
https://karpathy.ai/lexicap/0006-large.html#00:30:07.340
Guido van Rossum: Python | Lex Fridman Podcast #6
are going to do and they learn.
https://karpathy.ai/lexicap/0006-large.html#00:30:11.400
Guido van Rossum: Python | Lex Fridman Podcast #6
Some dogs know when you're going out and they want to go out with you, they're sad when
https://karpathy.ai/lexicap/0006-large.html#00:30:14.160
Guido van Rossum: Python | Lex Fridman Podcast #6
you leave them alone, they cry, they're afraid because they were mistreated when they were
https://karpathy.ai/lexicap/0006-large.html#00:30:19.060
Guido van Rossum: Python | Lex Fridman Podcast #6
younger.
https://karpathy.ai/lexicap/0006-large.html#00:30:26.080
Guido van Rossum: Python | Lex Fridman Podcast #6
We don't assign sort of consciousness to dogs, or at least not all that much, but I also
https://karpathy.ai/lexicap/0006-large.html#00:30:31.040
Guido van Rossum: Python | Lex Fridman Podcast #6
don't think they have none of that.
https://karpathy.ai/lexicap/0006-large.html#00:30:39.280
Guido van Rossum: Python | Lex Fridman Podcast #6
So I think it's consciousness and intelligence are not all or nothing.
https://karpathy.ai/lexicap/0006-large.html#00:30:42.500
Guido van Rossum: Python | Lex Fridman Podcast #6
The spectrum is really interesting.
https://karpathy.ai/lexicap/0006-large.html#00:30:50.360
Guido van Rossum: Python | Lex Fridman Podcast #6
But in returning to programming languages and the way we think about building these
https://karpathy.ai/lexicap/0006-large.html#00:30:52.780
Guido van Rossum: Python | Lex Fridman Podcast #6
kinds of things, about building intelligence, building consciousness, building artificial
https://karpathy.ai/lexicap/0006-large.html#00:30:58.760
Guido van Rossum: Python | Lex Fridman Podcast #6
beings.
https://karpathy.ai/lexicap/0006-large.html#00:31:03.260
Guido van Rossum: Python | Lex Fridman Podcast #6
So I think one of the exciting ideas came in the 17th century and with Leibniz, Hobbes,
https://karpathy.ai/lexicap/0006-large.html#00:31:04.260
Guido van Rossum: Python | Lex Fridman Podcast #6
Descartes, where there's this feeling that you can convert all thought, all reasoning,
https://karpathy.ai/lexicap/0006-large.html#00:31:10.920
Guido van Rossum: Python | Lex Fridman Podcast #6
all the thing that we find very special in our brains, you can convert all of that into
https://karpathy.ai/lexicap/0006-large.html#00:31:18.520
Guido van Rossum: Python | Lex Fridman Podcast #6
logic.
https://karpathy.ai/lexicap/0006-large.html#00:31:24.480
Guido van Rossum: Python | Lex Fridman Podcast #6
So you can formalize it, formal reasoning, and then once you formalize everything, all
https://karpathy.ai/lexicap/0006-large.html#00:31:25.480
Guido van Rossum: Python | Lex Fridman Podcast #6
of knowledge, then you can just calculate and that's what we're doing with our brains
https://karpathy.ai/lexicap/0006-large.html#00:31:30.400
Guido van Rossum: Python | Lex Fridman Podcast #6
is we're calculating.
https://karpathy.ai/lexicap/0006-large.html#00:31:34.400
Guido van Rossum: Python | Lex Fridman Podcast #6
So there's this whole idea that this is possible, that this we can actually program.
https://karpathy.ai/lexicap/0006-large.html#00:31:35.400
Guido van Rossum: Python | Lex Fridman Podcast #6
But they weren't aware of the concept of pattern matching in the sense that we are aware of
https://karpathy.ai/lexicap/0006-large.html#00:31:40.240
Guido van Rossum: Python | Lex Fridman Podcast #6
it now.
https://karpathy.ai/lexicap/0006-large.html#00:31:46.520
Guido van Rossum: Python | Lex Fridman Podcast #6
They sort of thought they had discovered incredible bits of mathematics like Newton's calculus
https://karpathy.ai/lexicap/0006-large.html#00:31:47.640
Guido van Rossum: Python | Lex Fridman Podcast #6
and their sort of idealism, their sort of extension of what they could do with logic
https://karpathy.ai/lexicap/0006-large.html#00:31:57.640
Guido van Rossum: Python | Lex Fridman Podcast #6
and math sort of went along those lines and they thought there's like, yeah, logic.
https://karpathy.ai/lexicap/0006-large.html#00:32:06.840
Guido van Rossum: Python | Lex Fridman Podcast #6
There's like a bunch of rules and a bunch of input.
https://karpathy.ai/lexicap/0006-large.html#00:32:18.000
Guido van Rossum: Python | Lex Fridman Podcast #6
They didn't realize that how you recognize a face is not just a bunch of rules but is
https://karpathy.ai/lexicap/0006-large.html#00:32:22.020
Guido van Rossum: Python | Lex Fridman Podcast #6
a shit ton of data plus a circuit that sort of interprets the visual clues and the context
https://karpathy.ai/lexicap/0006-large.html#00:32:28.600
Guido van Rossum: Python | Lex Fridman Podcast #6
and everything else and somehow can massively parallel pattern match against stored rules.
https://karpathy.ai/lexicap/0006-large.html#00:32:39.160
Guido van Rossum: Python | Lex Fridman Podcast #6
I mean, if I see you tomorrow here in front of the Dropbox office, I might recognize you.
https://karpathy.ai/lexicap/0006-large.html#00:32:49.400
Guido van Rossum: Python | Lex Fridman Podcast #6
Even if I'm wearing a different shirt, yeah, but if I see you tomorrow in a coffee shop
https://karpathy.ai/lexicap/0006-large.html#00:32:56.320
Guido van Rossum: Python | Lex Fridman Podcast #6
in Belmont, I might have no idea that it was you or on the beach or whatever.
https://karpathy.ai/lexicap/0006-large.html#00:33:01.320
Guido van Rossum: Python | Lex Fridman Podcast #6
I make those kind of mistakes myself all the time.
https://karpathy.ai/lexicap/0006-large.html#00:33:06.640
Guido van Rossum: Python | Lex Fridman Podcast #6
I see someone that I only know as like, oh, this person is a colleague of my wife's and
https://karpathy.ai/lexicap/0006-large.html#00:33:10.160
Guido van Rossum: Python | Lex Fridman Podcast #6
then I see them at the movies and I didn't recognize them.
https://karpathy.ai/lexicap/0006-large.html#00:33:16.320
Guido van Rossum: Python | Lex Fridman Podcast #6
But do you see those, you call it pattern matching, do you see that rules is unable
https://karpathy.ai/lexicap/0006-large.html#00:33:20.860
Guido van Rossum: Python | Lex Fridman Podcast #6
to encode that?
https://karpathy.ai/lexicap/0006-large.html#00:33:29.320
Guido van Rossum: Python | Lex Fridman Podcast #6
Everything you see, all the pieces of information you look around this room, I'm wearing a black
https://karpathy.ai/lexicap/0006-large.html#00:33:32.380
Guido van Rossum: Python | Lex Fridman Podcast #6
shirt, I have a certain height, I'm a human, all these, there's probably tens of thousands
https://karpathy.ai/lexicap/0006-large.html#00:33:36.320
Guido van Rossum: Python | Lex Fridman Podcast #6
of facts you pick up moment by moment about this scene.
https://karpathy.ai/lexicap/0006-large.html#00:33:41.720
Guido van Rossum: Python | Lex Fridman Podcast #6
You take them for granted and you aggregate them together to understand the scene.
https://karpathy.ai/lexicap/0006-large.html#00:33:45.680
Guido van Rossum: Python | Lex Fridman Podcast #6
You don't think all of that could be encoded to where at the end of the day, you can just
https://karpathy.ai/lexicap/0006-large.html#00:33:50.000
Guido van Rossum: Python | Lex Fridman Podcast #6
put it all on the table and calculate?
https://karpathy.ai/lexicap/0006-large.html#00:33:53.800
Guido van Rossum: Python | Lex Fridman Podcast #6
I don't know what that means.
https://karpathy.ai/lexicap/0006-large.html#00:33:57.440
Guido van Rossum: Python | Lex Fridman Podcast #6
I mean, yes, in the sense that there is no actual magic there, but there are enough layers
https://karpathy.ai/lexicap/0006-large.html#00:33:58.840
Guido van Rossum: Python | Lex Fridman Podcast #6
of abstraction from the facts as they enter my eyes and my ears to the understanding of
https://karpathy.ai/lexicap/0006-large.html#00:34:08.680
Guido van Rossum: Python | Lex Fridman Podcast #6
the scene that I don't think that AI has really covered enough of that distance.
https://karpathy.ai/lexicap/0006-large.html#00:34:17.640
Guido van Rossum: Python | Lex Fridman Podcast #6
It's like if you take a human body and you realize it's built out of atoms, well, that
https://karpathy.ai/lexicap/0006-large.html#00:34:29.240
Guido van Rossum: Python | Lex Fridman Podcast #6
is a uselessly reductionist view, right?
https://karpathy.ai/lexicap/0006-large.html#00:34:37.800
Guido van Rossum: Python | Lex Fridman Podcast #6
The body is built out of organs, the organs are built out of cells, the cells are built
https://karpathy.ai/lexicap/0006-large.html#00:34:41.960
Guido van Rossum: Python | Lex Fridman Podcast #6
out of proteins, the proteins are built out of amino acids, the amino acids are built
https://karpathy.ai/lexicap/0006-large.html#00:34:46.380
Guido van Rossum: Python | Lex Fridman Podcast #6
out of atoms and then you get to quantum mechanics.
https://karpathy.ai/lexicap/0006-large.html#00:34:53.240
Guido van Rossum: Python | Lex Fridman Podcast #6
So that's a very pragmatic view.
https://karpathy.ai/lexicap/0006-large.html#00:34:58.040
Guido van Rossum: Python | Lex Fridman Podcast #6
I mean, obviously as an engineer, I agree with that kind of view, but you also have
https://karpathy.ai/lexicap/0006-large.html#00:34:59.920
Guido van Rossum: Python | Lex Fridman Podcast #6
to consider the Sam Harris view of, well, intelligence is just information processing.
https://karpathy.ai/lexicap/0006-large.html#00:35:03.720
Guido van Rossum: Python | Lex Fridman Podcast #6
Like you said, you take in sensory information, you do some stuff with it and you come up
https://karpathy.ai/lexicap/0006-large.html#00:35:13.160
Guido van Rossum: Python | Lex Fridman Podcast #6
with actions that are intelligent.
https://karpathy.ai/lexicap/0006-large.html#00:35:17.320
Guido van Rossum: Python | Lex Fridman Podcast #6
That makes it sound so easy.
https://karpathy.ai/lexicap/0006-large.html#00:35:20.760
Guido van Rossum: Python | Lex Fridman Podcast #6
I don't know who Sam Harris is.
https://karpathy.ai/lexicap/0006-large.html#00:35:22.480
Guido van Rossum: Python | Lex Fridman Podcast #6
Oh, well, it's a philosopher.
https://karpathy.ai/lexicap/0006-large.html#00:35:24.280
Guido van Rossum: Python | Lex Fridman Podcast #6
So like this is how philosophers often think, right?
https://karpathy.ai/lexicap/0006-large.html#00:35:26.400
Guido van Rossum: Python | Lex Fridman Podcast #6
And essentially that's what Descartes was, is wait a minute, if there is, like you said,
https://karpathy.ai/lexicap/0006-large.html#00:35:29.680
Guido van Rossum: Python | Lex Fridman Podcast #6
no magic, so he basically says it doesn't appear like there's any magic, but we know
https://karpathy.ai/lexicap/0006-large.html#00:35:33.760
Guido van Rossum: Python | Lex Fridman Podcast #6
so little about it that it might as well be magic.
https://karpathy.ai/lexicap/0006-large.html#00:35:39.320
Guido van Rossum: Python | Lex Fridman Podcast #6
So just because we know that we're made of atoms, just because we know we're made
https://karpathy.ai/lexicap/0006-large.html#00:35:44.280
Guido van Rossum: Python | Lex Fridman Podcast #6
of organs, the fact that we know very little how to get from the atoms to organs in a way
https://karpathy.ai/lexicap/0006-large.html#00:35:47.800
Guido van Rossum: Python | Lex Fridman Podcast #6
that's recreatable means that you shouldn't get too excited just yet about the fact that
https://karpathy.ai/lexicap/0006-large.html#00:35:53.280
Guido van Rossum: Python | Lex Fridman Podcast #6
you figured out that we're made of atoms.
https://karpathy.ai/lexicap/0006-large.html#00:36:00.400
Guido van Rossum: Python | Lex Fridman Podcast #6
Right, and the same about taking facts as our sensory organs take them in and turning
https://karpathy.ai/lexicap/0006-large.html#00:36:02.240
Guido van Rossum: Python | Lex Fridman Podcast #6
that into reasons and actions, that sort of, there are a lot of abstractions that we haven't
https://karpathy.ai/lexicap/0006-large.html#00:36:11.920
Guido van Rossum: Python | Lex Fridman Podcast #6
quite figured out how to deal with those.
https://karpathy.ai/lexicap/0006-large.html#00:36:19.820
Guido van Rossum: Python | Lex Fridman Podcast #6
I mean, sometimes, I don't know if I can go on a tangent or not, so if I take a simple
https://karpathy.ai/lexicap/0006-large.html#00:36:23.960
Guido van Rossum: Python | Lex Fridman Podcast #6
program that parses, say I have a compiler that parses a program, in a sense the input
https://karpathy.ai/lexicap/0006-large.html#00:36:37.440
Guido van Rossum: Python | Lex Fridman Podcast #6
routine of that compiler, of that parser, is a sensing organ, and it builds up a mighty
https://karpathy.ai/lexicap/0006-large.html#00:36:45.640
Guido van Rossum: Python | Lex Fridman Podcast #6
complicated internal representation of the program it just saw, it doesn't just have
https://karpathy.ai/lexicap/0006-large.html#00:36:55.640
Guido van Rossum: Python | Lex Fridman Podcast #6
a linear sequence of bytes representing the text of the program anymore, it has an abstract
https://karpathy.ai/lexicap/0006-large.html#00:37:01.960
Guido van Rossum: Python | Lex Fridman Podcast #6
syntax tree, and I don't know how many of your viewers or listeners are familiar with
https://karpathy.ai/lexicap/0006-large.html#00:37:08.200
Guido van Rossum: Python | Lex Fridman Podcast #6
compiler technology, but there's…
https://karpathy.ai/lexicap/0006-large.html#00:37:15.480
Guido van Rossum: Python | Lex Fridman Podcast #6
Fewer and fewer these days, right?
https://karpathy.ai/lexicap/0006-large.html#00:37:18.680
Guido van Rossum: Python | Lex Fridman Podcast #6
That's also true, probably.
https://karpathy.ai/lexicap/0006-large.html#00:37:21.880
Guido van Rossum: Python | Lex Fridman Podcast #6
People want to take a shortcut, but there's sort of, this abstraction is a data structure
https://karpathy.ai/lexicap/0006-large.html#00:37:24.920
Guido van Rossum: Python | Lex Fridman Podcast #6
that the compiler then uses to produce outputs that is relevant, like a translation of that
https://karpathy.ai/lexicap/0006-large.html#00:37:30.360
Guido van Rossum: Python | Lex Fridman Podcast #6
program to machine code that can be executed by hardware, and then that data structure
https://karpathy.ai/lexicap/0006-large.html#00:37:37.480
Guido van Rossum: Python | Lex Fridman Podcast #6
gets thrown away.
https://karpathy.ai/lexicap/0006-large.html#00:37:47.880
Guido van Rossum: Python | Lex Fridman Podcast #6
When a fish or a fly sees, sort of gets visual impulses, I'm sure it also builds up some
https://karpathy.ai/lexicap/0006-large.html#00:37:50.600
Guido van Rossum: Python | Lex Fridman Podcast #6
data structure, and for the fly that may be very minimal, a fly may have only a few, I
https://karpathy.ai/lexicap/0006-large.html#00:38:02.560
Guido van Rossum: Python | Lex Fridman Podcast #6
mean, in the case of a fly's brain, I could imagine that there are few enough layers of
https://karpathy.ai/lexicap/0006-large.html#00:38:10.000
Guido van Rossum: Python | Lex Fridman Podcast #6
abstraction that it's not much more than when it's darker here than it is here, well
https://karpathy.ai/lexicap/0006-large.html#00:38:17.680
Guido van Rossum: Python | Lex Fridman Podcast #6
it can sense motion, because a fly sort of responds when you move your arm towards it,
https://karpathy.ai/lexicap/0006-large.html#00:38:24.040
Guido van Rossum: Python | Lex Fridman Podcast #6
so clearly its visual processing is intelligent, well, not intelligent, but it has an abstraction
https://karpathy.ai/lexicap/0006-large.html#00:38:29.880
Guido van Rossum: Python | Lex Fridman Podcast #6
for motion, and we still have similar things in, but much more complicated in our brains,
https://karpathy.ai/lexicap/0006-large.html#00:38:39.240
Guido van Rossum: Python | Lex Fridman Podcast #6
I mean, otherwise you couldn't drive a car if you couldn't, if you didn't have an
https://karpathy.ai/lexicap/0006-large.html#00:38:46.440
Guido van Rossum: Python | Lex Fridman Podcast #6
incredibly good abstraction for motion.
https://karpathy.ai/lexicap/0006-large.html#00:38:50.400
Guido van Rossum: Python | Lex Fridman Podcast #6
Yeah, in some sense, the same abstraction for motion is probably one of the primary
https://karpathy.ai/lexicap/0006-large.html#00:38:53.480
Guido van Rossum: Python | Lex Fridman Podcast #6
sources of our, of information for us, we just know what to do, I think we know what
https://karpathy.ai/lexicap/0006-large.html#00:38:59.160
Guido van Rossum: Python | Lex Fridman Podcast #6
to do with that, we've built up other abstractions on top.
https://karpathy.ai/lexicap/0006-large.html#00:39:05.080
Guido van Rossum: Python | Lex Fridman Podcast #6
We build much more complicated data structures based on that, and we build more persistent
https://karpathy.ai/lexicap/0006-large.html#00:39:08.280
Guido van Rossum: Python | Lex Fridman Podcast #6
data structures, sort of after some processing, some information sort of gets stored in our
https://karpathy.ai/lexicap/0006-large.html#00:39:14.120
Guido van Rossum: Python | Lex Fridman Podcast #6
memory pretty much permanently, and is available on recall, I mean, there are some things that
https://karpathy.ai/lexicap/0006-large.html#00:39:20.320
Guido van Rossum: Python | Lex Fridman Podcast #6
you sort of, you're conscious that you're remembering it, like, you give me your phone
https://karpathy.ai/lexicap/0006-large.html#00:39:27.240