episode
stringlengths
45
100
text
stringlengths
1
528
timestamp_link
stringlengths
56
56
Guido van Rossum: Python | Lex Fridman Podcast #6
education and a different model of what they're trying to do will develop those concepts.
https://karpathy.ai/lexicap/0006-large.html#00:48:56.200
Guido van Rossum: Python | Lex Fridman Podcast #6
I guess if you could comment on another way to put this concept is, I think the kind of
https://karpathy.ai/lexicap/0006-large.html#00:49:07.920
Guido van Rossum: Python | Lex Fridman Podcast #6
functions that neural networks provide is things as opposed to being able to upfront
https://karpathy.ai/lexicap/0006-large.html#00:49:17.240
Guido van Rossum: Python | Lex Fridman Podcast #6
prove that this should work for all cases you throw at it.
https://karpathy.ai/lexicap/0006-large.html#00:49:23.480
Guido van Rossum: Python | Lex Fridman Podcast #6
All you're able, it's the worst case analysis versus average case analysis.
https://karpathy.ai/lexicap/0006-large.html#00:49:28.720
Guido van Rossum: Python | Lex Fridman Podcast #6
All you're able to say is it seems on everything we've tested to work 99.9% of the time, but
https://karpathy.ai/lexicap/0006-large.html#00:49:32.320
Guido van Rossum: Python | Lex Fridman Podcast #6
we can't guarantee it and it fails in unexpected ways.
https://karpathy.ai/lexicap/0006-large.html#00:49:39.800
Guido van Rossum: Python | Lex Fridman Podcast #6
We can't even give you examples of how it fails in unexpected ways, but it's like really
https://karpathy.ai/lexicap/0006-large.html#00:49:44.160
Guido van Rossum: Python | Lex Fridman Podcast #6
good most of the time.
https://karpathy.ai/lexicap/0006-large.html#00:49:48.080
Guido van Rossum: Python | Lex Fridman Podcast #6
Is there no room for that in current ways we think about programming?
https://karpathy.ai/lexicap/0006-large.html#00:49:50.120
Guido van Rossum: Python | Lex Fridman Podcast #6
programming 1.0 is actually sort of getting to that point too, where the sort of the ideal
https://karpathy.ai/lexicap/0006-large.html#00:50:00.720
Guido van Rossum: Python | Lex Fridman Podcast #6
of a bug free program has been abandoned long ago by most software developers.
https://karpathy.ai/lexicap/0006-large.html#00:50:11.080
Guido van Rossum: Python | Lex Fridman Podcast #6
We only care about bugs that manifest themselves often enough to be annoying.
https://karpathy.ai/lexicap/0006-large.html#00:50:21.120
Guido van Rossum: Python | Lex Fridman Podcast #6
And we're willing to take the occasional crash or outage or incorrect result for granted
https://karpathy.ai/lexicap/0006-large.html#00:50:30.120
Guido van Rossum: Python | Lex Fridman Podcast #6
because we can't possibly, we don't have enough programmers to make all the code bug free
https://karpathy.ai/lexicap/0006-large.html#00:50:40.680
Guido van Rossum: Python | Lex Fridman Podcast #6
and it would be an incredibly tedious business.
https://karpathy.ai/lexicap/0006-large.html#00:50:47.600
Guido van Rossum: Python | Lex Fridman Podcast #6
And if you try to throw formal methods at it, it becomes even more tedious.
https://karpathy.ai/lexicap/0006-large.html#00:50:50.200
Guido van Rossum: Python | Lex Fridman Podcast #6
So every once in a while the user clicks on a link and somehow they get an error and the
https://karpathy.ai/lexicap/0006-large.html#00:50:56.320
Guido van Rossum: Python | Lex Fridman Podcast #6
average user doesn't panic.
https://karpathy.ai/lexicap/0006-large.html#00:51:05.520
Guido van Rossum: Python | Lex Fridman Podcast #6
They just click again and see if it works better the second time, which often magically
https://karpathy.ai/lexicap/0006-large.html#00:51:07.360
Guido van Rossum: Python | Lex Fridman Podcast #6
it does, or they go up and they try some other way of performing their tasks.
https://karpathy.ai/lexicap/0006-large.html#00:51:14.840
Guido van Rossum: Python | Lex Fridman Podcast #6
So that's sort of an end to end recovery mechanism and inside systems there is all
https://karpathy.ai/lexicap/0006-large.html#00:51:21.600
Guido van Rossum: Python | Lex Fridman Podcast #6
sorts of retries and timeouts and fallbacks and I imagine that that sort of biological
https://karpathy.ai/lexicap/0006-large.html#00:51:29.880
Guido van Rossum: Python | Lex Fridman Podcast #6
systems are even more full of that because otherwise they wouldn't survive.
https://karpathy.ai/lexicap/0006-large.html#00:51:39.120
Guido van Rossum: Python | Lex Fridman Podcast #6
Do you think programming should be taught and thought of as exactly what you just said?
https://karpathy.ai/lexicap/0006-large.html#00:51:46.320
Guido van Rossum: Python | Lex Fridman Podcast #6
I come from this kind of, you're always denying that fact always.
https://karpathy.ai/lexicap/0006-large.html#00:51:54.160
Guido van Rossum: Python | Lex Fridman Podcast #6
In sort of basic programming education, the sort of the programs you're having students
https://karpathy.ai/lexicap/0006-large.html#00:52:01.560
Guido van Rossum: Python | Lex Fridman Podcast #6
write are so small and simple that if there is a bug you can always find it and fix it.
https://karpathy.ai/lexicap/0006-large.html#00:52:12.680
Guido van Rossum: Python | Lex Fridman Podcast #6
Because the sort of programming as it's being taught in some, even elementary, middle schools,
https://karpathy.ai/lexicap/0006-large.html#00:52:23.480
Guido van Rossum: Python | Lex Fridman Podcast #6
in high school, introduction to programming classes in college typically, it's programming
https://karpathy.ai/lexicap/0006-large.html#00:52:29.720
Guido van Rossum: Python | Lex Fridman Podcast #6
in the small.
https://karpathy.ai/lexicap/0006-large.html#00:52:36.680
Guido van Rossum: Python | Lex Fridman Podcast #6
Very few classes sort of actually teach software engineering, building large systems.
https://karpathy.ai/lexicap/0006-large.html#00:52:38.920
Guido van Rossum: Python | Lex Fridman Podcast #6
Every summer here at Dropbox we have a large number of interns.
https://karpathy.ai/lexicap/0006-large.html#00:52:47.560
Guido van Rossum: Python | Lex Fridman Podcast #6
Every tech company on the West Coast has the same thing.
https://karpathy.ai/lexicap/0006-large.html#00:52:51.360
Guido van Rossum: Python | Lex Fridman Podcast #6
These interns are always amazed because this is the first time in their life that they
https://karpathy.ai/lexicap/0006-large.html#00:52:56.720
Guido van Rossum: Python | Lex Fridman Podcast #6
see what goes on in a really large software development environment.
https://karpathy.ai/lexicap/0006-large.html#00:53:02.520
Guido van Rossum: Python | Lex Fridman Podcast #6
Everything they've learned in college was almost always about a much smaller scale and
https://karpathy.ai/lexicap/0006-large.html#00:53:12.920
Guido van Rossum: Python | Lex Fridman Podcast #6
somehow that difference in scale makes a qualitative difference in how you do things and how you
https://karpathy.ai/lexicap/0006-large.html#00:53:20.280
Guido van Rossum: Python | Lex Fridman Podcast #6
think about it.
https://karpathy.ai/lexicap/0006-large.html#00:53:27.840
Guido van Rossum: Python | Lex Fridman Podcast #6
If you then take a few steps back into decades, 70s and 80s, when you were first thinking
https://karpathy.ai/lexicap/0006-large.html#00:53:29.600
Guido van Rossum: Python | Lex Fridman Podcast #6
about Python or just that world of programming languages, did you ever think that there would
https://karpathy.ai/lexicap/0006-large.html#00:53:36.300
Guido van Rossum: Python | Lex Fridman Podcast #6
be systems as large as underlying Google, Facebook, and Dropbox?
https://karpathy.ai/lexicap/0006-large.html#00:53:41.840
Guido van Rossum: Python | Lex Fridman Podcast #6
Did you, when you were thinking about Python?
https://karpathy.ai/lexicap/0006-large.html#00:53:46.720
Guido van Rossum: Python | Lex Fridman Podcast #6
I was actually always caught by surprise by sort of this, yeah, pretty much every stage
https://karpathy.ai/lexicap/0006-large.html#00:53:51.440
Guido van Rossum: Python | Lex Fridman Podcast #6
of computing.
https://karpathy.ai/lexicap/0006-large.html#00:53:57.520
Guido van Rossum: Python | Lex Fridman Podcast #6
So maybe just because you've spoken in other interviews, but I think the evolution of programming
https://karpathy.ai/lexicap/0006-large.html#00:53:59.680
Guido van Rossum: Python | Lex Fridman Podcast #6
languages are fascinating and it's especially because it leads from my perspective towards
https://karpathy.ai/lexicap/0006-large.html#00:54:07.280
Guido van Rossum: Python | Lex Fridman Podcast #6
greater and greater degrees of intelligence.
https://karpathy.ai/lexicap/0006-large.html#00:54:13.080
Guido van Rossum: Python | Lex Fridman Podcast #6
I learned the first programming language I played with in Russia was with the Turtle
https://karpathy.ai/lexicap/0006-large.html#00:54:15.640
Guido van Rossum: Python | Lex Fridman Podcast #6
logo.
https://karpathy.ai/lexicap/0006-large.html#00:54:21.880
Guido van Rossum: Python | Lex Fridman Podcast #6
Logo, yeah.
https://karpathy.ai/lexicap/0006-large.html#00:54:22.880
Guido van Rossum: Python | Lex Fridman Podcast #6
And if you look, I just have a list of programming languages, all of which I've now played with
https://karpathy.ai/lexicap/0006-large.html#00:54:24.840
Guido van Rossum: Python | Lex Fridman Podcast #6
a little bit.
https://karpathy.ai/lexicap/0006-large.html#00:54:29.960
Guido van Rossum: Python | Lex Fridman Podcast #6
I mean, they're all beautiful in different ways from Fortran, Cobalt, Lisp, Algol 60,
https://karpathy.ai/lexicap/0006-large.html#00:54:30.960
Guido van Rossum: Python | Lex Fridman Podcast #6
Basic, Logo again, C, as a few, the object oriented came along in the 60s, Simula, Pascal,
https://karpathy.ai/lexicap/0006-large.html#00:54:36.640
Guido van Rossum: Python | Lex Fridman Podcast #6
Smalltalk.
https://karpathy.ai/lexicap/0006-large.html#00:54:46.160
Guido van Rossum: Python | Lex Fridman Podcast #6
All of that leads.
https://karpathy.ai/lexicap/0006-large.html#00:54:47.560
Guido van Rossum: Python | Lex Fridman Podcast #6
They're all the classics.
https://karpathy.ai/lexicap/0006-large.html#00:54:48.560
Guido van Rossum: Python | Lex Fridman Podcast #6
The classics.
https://karpathy.ai/lexicap/0006-large.html#00:54:49.560
Guido van Rossum: Python | Lex Fridman Podcast #6
Yeah.
https://karpathy.ai/lexicap/0006-large.html#00:54:50.560
Guido van Rossum: Python | Lex Fridman Podcast #6
The classic hits, right?
https://karpathy.ai/lexicap/0006-large.html#00:54:51.560
Guido van Rossum: Python | Lex Fridman Podcast #6
Steam, that's built on top of Lisp.
https://karpathy.ai/lexicap/0006-large.html#00:54:52.560
Guido van Rossum: Python | Lex Fridman Podcast #6
On the database side, SQL, C++, and all of that leads up to Python, Pascal too, and that's
https://karpathy.ai/lexicap/0006-large.html#00:54:58.280
Guido van Rossum: Python | Lex Fridman Podcast #6
before Python, MATLAB, these kind of different communities, different languages.
https://karpathy.ai/lexicap/0006-large.html#00:55:05.900
Guido van Rossum: Python | Lex Fridman Podcast #6
So can you talk about that world?
https://karpathy.ai/lexicap/0006-large.html#00:55:10.960
Guido van Rossum: Python | Lex Fridman Podcast #6
I know that sort of Python came out of ABC, which I actually never knew that language.
https://karpathy.ai/lexicap/0006-large.html#00:55:13.240
Guido van Rossum: Python | Lex Fridman Podcast #6
I just, having researched this conversation, went back to ABC and it looks remarkably,
https://karpathy.ai/lexicap/0006-large.html#00:55:18.680
Guido van Rossum: Python | Lex Fridman Podcast #6
it has a lot of annoying qualities, but underneath those, like all caps and so on, but underneath
https://karpathy.ai/lexicap/0006-large.html#00:55:24.400
Guido van Rossum: Python | Lex Fridman Podcast #6
that, there's elements of Python that are quite, they're already there.
https://karpathy.ai/lexicap/0006-large.html#00:55:31.240
Guido van Rossum: Python | Lex Fridman Podcast #6
That's where I got all the good stuff.
https://karpathy.ai/lexicap/0006-large.html#00:55:35.720
Guido van Rossum: Python | Lex Fridman Podcast #6
All the good stuff.
https://karpathy.ai/lexicap/0006-large.html#00:55:37.540
Guido van Rossum: Python | Lex Fridman Podcast #6
So, but in that world, you're swimming these programming languages, were you focused on
https://karpathy.ai/lexicap/0006-large.html#00:55:38.540
Guido van Rossum: Python | Lex Fridman Podcast #6
just the good stuff in your specific circle, or did you have a sense of what is everyone
https://karpathy.ai/lexicap/0006-large.html#00:55:41.580
Guido van Rossum: Python | Lex Fridman Podcast #6
chasing?
https://karpathy.ai/lexicap/0006-large.html#00:55:48.080
Guido van Rossum: Python | Lex Fridman Podcast #6
You said that every programming language is built to scratch an itch.
https://karpathy.ai/lexicap/0006-large.html#00:55:49.080
Guido van Rossum: Python | Lex Fridman Podcast #6
Were you aware of all the itches in the community?
https://karpathy.ai/lexicap/0006-large.html#00:55:57.000
Guido van Rossum: Python | Lex Fridman Podcast #6
And if not, or if yes, I mean, what itch were you trying to scratch with Python?
https://karpathy.ai/lexicap/0006-large.html#00:55:59.920
Guido van Rossum: Python | Lex Fridman Podcast #6
Well, I'm glad I wasn't aware of all the itches because I would probably not have been able
https://karpathy.ai/lexicap/0006-large.html#00:56:05.080
Guido van Rossum: Python | Lex Fridman Podcast #6
to do anything.
https://karpathy.ai/lexicap/0006-large.html#00:56:12.040
Guido van Rossum: Python | Lex Fridman Podcast #6
I mean, if you're trying to solve every problem at once, you'll solve nothing.
https://karpathy.ai/lexicap/0006-large.html#00:56:14.040
Guido van Rossum: Python | Lex Fridman Podcast #6
Well, yeah, it's too overwhelming.
https://karpathy.ai/lexicap/0006-large.html#00:56:19.760
Guido van Rossum: Python | Lex Fridman Podcast #6
And so I had a very, very focused problem.
https://karpathy.ai/lexicap/0006-large.html#00:56:23.880
Guido van Rossum: Python | Lex Fridman Podcast #6
I wanted a programming language that sat somewhere in between shell scripting and C. And now,
https://karpathy.ai/lexicap/0006-large.html#00:56:28.360
Guido van Rossum: Python | Lex Fridman Podcast #6
arguably, there is like, one is higher level, one is lower level.
https://karpathy.ai/lexicap/0006-large.html#00:56:41.880
Guido van Rossum: Python | Lex Fridman Podcast #6
And Python is sort of a language of an intermediate level, although it's still pretty much at
https://karpathy.ai/lexicap/0006-large.html#00:56:48.720
Guido van Rossum: Python | Lex Fridman Podcast #6
the high level end.
https://karpathy.ai/lexicap/0006-large.html#00:56:56.760
Guido van Rossum: Python | Lex Fridman Podcast #6
I was thinking about much more about, I want a tool that I can use to be more productive
https://karpathy.ai/lexicap/0006-large.html#00:57:00.560
Guido van Rossum: Python | Lex Fridman Podcast #6
as a programmer in a very specific environment.
https://karpathy.ai/lexicap/0006-large.html#00:57:11.200
Guido van Rossum: Python | Lex Fridman Podcast #6
And I also had given myself a time budget for the development of the tool.
https://karpathy.ai/lexicap/0006-large.html#00:57:16.640
Guido van Rossum: Python | Lex Fridman Podcast #6
And that was sort of about three months for both the design, like thinking through what
https://karpathy.ai/lexicap/0006-large.html#00:57:22.280
Guido van Rossum: Python | Lex Fridman Podcast #6
are all the features of the language syntactically and semantically, and how do I implement the
https://karpathy.ai/lexicap/0006-large.html#00:57:29.340
Guido van Rossum: Python | Lex Fridman Podcast #6
whole pipeline from parsing the source code to executing it.
https://karpathy.ai/lexicap/0006-large.html#00:57:38.900
Guido van Rossum: Python | Lex Fridman Podcast #6
So I think both with the timeline and the goals, it seems like productivity was at the
https://karpathy.ai/lexicap/0006-large.html#00:57:43.680
Guido van Rossum: Python | Lex Fridman Podcast #6
core of it as a goal.
https://karpathy.ai/lexicap/0006-large.html#00:57:51.440
Guido van Rossum: Python | Lex Fridman Podcast #6
So like, for me in the 90s, and the first decade of the 21st century, I was always doing
https://karpathy.ai/lexicap/0006-large.html#00:57:54.040
Guido van Rossum: Python | Lex Fridman Podcast #6
machine learning, AI programming for my research was always in C++.
https://karpathy.ai/lexicap/0006-large.html#00:58:01.280
Guido van Rossum: Python | Lex Fridman Podcast #6
And then the other people who are a little more mechanical engineering, electrical engineering,
https://karpathy.ai/lexicap/0006-large.html#00:58:07.620
Guido van Rossum: Python | Lex Fridman Podcast #6
are MATLABby.
https://karpathy.ai/lexicap/0006-large.html#00:58:14.240
Guido van Rossum: Python | Lex Fridman Podcast #6
They're a little bit more MATLAB focused.
https://karpathy.ai/lexicap/0006-large.html#00:58:15.240
Guido van Rossum: Python | Lex Fridman Podcast #6
Those are the world, and maybe a little bit Java too.
https://karpathy.ai/lexicap/0006-large.html#00:58:18.520