episode
stringlengths
45
100
text
stringlengths
1
528
timestamp_link
stringlengths
56
56
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
these transportation optimization systems.
https://karpathy.ai/lexicap/0012-large.html#00:57:01.560
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And somebody had heard that it's a good idea
https://karpathy.ai/lexicap/0012-large.html#00:57:03.560
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
to have high utilization of assets.
https://karpathy.ai/lexicap/0012-large.html#00:57:05.360
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So they told me, hey, why don't you put that as objective?
https://karpathy.ai/lexicap/0012-large.html#00:57:08.160
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And we didn't even put it as an objective
https://karpathy.ai/lexicap/0012-large.html#00:57:11.400
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
because I just showed him that,
https://karpathy.ai/lexicap/0012-large.html#00:57:14.720
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
if you had that as your objective,
https://karpathy.ai/lexicap/0012-large.html#00:57:16.480
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
the solution would be to load your trucks full
https://karpathy.ai/lexicap/0012-large.html#00:57:18.480
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
and drive in circles.
https://karpathy.ai/lexicap/0012-large.html#00:57:20.320
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Nothing would ever get delivered.
https://karpathy.ai/lexicap/0012-large.html#00:57:21.840
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
You'd have a hundred percent utilization.
https://karpathy.ai/lexicap/0012-large.html#00:57:23.000
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So yeah, I know this phenomenon.
https://karpathy.ai/lexicap/0012-large.html#00:57:25.120
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
I've known this for over 30 years,
https://karpathy.ai/lexicap/0012-large.html#00:57:27.240
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
but I've never seen it actually be a problem in reality.
https://karpathy.ai/lexicap/0012-large.html#00:57:29.680
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And yes, if you have the wrong objective,
https://karpathy.ai/lexicap/0012-large.html#00:57:33.360
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
the AI will optimize that to the hilt
https://karpathy.ai/lexicap/0012-large.html#00:57:35.240
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
and it's gonna hurt more than some human
https://karpathy.ai/lexicap/0012-large.html#00:57:37.800
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
who's kind of trying to solve it in a half baked way
https://karpathy.ai/lexicap/0012-large.html#00:57:39.800
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
with some human insight too.
https://karpathy.ai/lexicap/0012-large.html#00:57:43.800
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
But I just haven't seen that materialize in practice.
https://karpathy.ai/lexicap/0012-large.html#00:57:45.480
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
There's this gap that you've actually put your finger on
https://karpathy.ai/lexicap/0012-large.html#00:57:49.160
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
very clearly just now between theory and reality.
https://karpathy.ai/lexicap/0012-large.html#00:57:52.720
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
That's very difficult to put into words, I think.
https://karpathy.ai/lexicap/0012-large.html#00:57:57.080
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
It's what you can theoretically imagine,
https://karpathy.ai/lexicap/0012-large.html#00:57:59.680
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
the worst possible case or even, yeah, I mean bad cases.
https://karpathy.ai/lexicap/0012-large.html#00:58:03.240
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And what usually happens in reality.
https://karpathy.ai/lexicap/0012-large.html#00:58:08.000
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So for example, to me,
https://karpathy.ai/lexicap/0012-large.html#00:58:10.520
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
maybe it's something you can comment on having grown up
https://karpathy.ai/lexicap/0012-large.html#00:58:11.960
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
and I grew up in the Soviet Union.
https://karpathy.ai/lexicap/0012-large.html#00:58:15.720
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
There's currently 10,000 nuclear weapons in the world.
https://karpathy.ai/lexicap/0012-large.html#00:58:19.120
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And for many decades,
https://karpathy.ai/lexicap/0012-large.html#00:58:22.120
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
it's theoretically surprising to me
https://karpathy.ai/lexicap/0012-large.html#00:58:24.200
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
that the nuclear war is not broken out.
https://karpathy.ai/lexicap/0012-large.html#00:58:28.360
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Do you think about this aspect
https://karpathy.ai/lexicap/0012-large.html#00:58:30.880
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
from a game theoretic perspective in general,
https://karpathy.ai/lexicap/0012-large.html#00:58:33.760
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
why is that true?
https://karpathy.ai/lexicap/0012-large.html#00:58:36.080
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Why in theory you could see
https://karpathy.ai/lexicap/0012-large.html#00:58:38.440
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
how things would go terribly wrong
https://karpathy.ai/lexicap/0012-large.html#00:58:40.720
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
and somehow yet they have not?
https://karpathy.ai/lexicap/0012-large.html#00:58:42.600
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Yeah, how do you think about it?
https://karpathy.ai/lexicap/0012-large.html#00:58:44.280
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So I do think about that a lot.
https://karpathy.ai/lexicap/0012-large.html#00:58:45.600
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
I think the biggest two threats that we're facing as mankind,
https://karpathy.ai/lexicap/0012-large.html#00:58:47.240
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
one is climate change and the other is nuclear war.
https://karpathy.ai/lexicap/0012-large.html#00:58:50.320
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So those are my main two worries that I worry about.
https://karpathy.ai/lexicap/0012-large.html#00:58:53.320
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And I've tried to do something about climate,
https://karpathy.ai/lexicap/0012-large.html#00:58:57.200
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
thought about trying to do something
https://karpathy.ai/lexicap/0012-large.html#00:58:59.920
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
for climate change twice.
https://karpathy.ai/lexicap/0012-large.html#00:59:01.320
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Actually, for two of my startups,
https://karpathy.ai/lexicap/0012-large.html#00:59:02.880
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
I've actually commissioned studies
https://karpathy.ai/lexicap/0012-large.html#00:59:05.040
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
of what we could do on those things.
https://karpathy.ai/lexicap/0012-large.html#00:59:06.760
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And we didn't really find a sweet spot,
https://karpathy.ai/lexicap/0012-large.html#00:59:09.480
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
but I'm still keeping an eye out on that.
https://karpathy.ai/lexicap/0012-large.html#00:59:11.040
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
If there's something where we could actually
https://karpathy.ai/lexicap/0012-large.html#00:59:12.680
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
provide a market solution or optimizations solution
https://karpathy.ai/lexicap/0012-large.html#00:59:15.160
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
or some other technology solution to problems.
https://karpathy.ai/lexicap/0012-large.html#00:59:17.800
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Right now, like for example,
https://karpathy.ai/lexicap/0012-large.html#00:59:20.960
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
pollution critic markets was what we were looking at then.
https://karpathy.ai/lexicap/0012-large.html#00:59:23.360
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And it was much more the lack of political will
https://karpathy.ai/lexicap/0012-large.html#00:59:26.760
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
by those markets were not so successful
https://karpathy.ai/lexicap/0012-large.html#00:59:30.040
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
rather than bad market design.
https://karpathy.ai/lexicap/0012-large.html#00:59:32.840
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So I could go in and make a better market design,
https://karpathy.ai/lexicap/0012-large.html#00:59:34.640
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
but that wouldn't really move the needle
https://karpathy.ai/lexicap/0012-large.html#00:59:37.080
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
on the world very much if there's no political will.
https://karpathy.ai/lexicap/0012-large.html#00:59:38.600
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And in the US, the market,
https://karpathy.ai/lexicap/0012-large.html#00:59:41.160
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
at least the Chicago market was just shut down and so on.
https://karpathy.ai/lexicap/0012-large.html#00:59:43.600
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So then it doesn't really help
https://karpathy.ai/lexicap/0012-large.html#00:59:47.520
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
how great your market design was.
https://karpathy.ai/lexicap/0012-large.html#00:59:48.760
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And then the nuclear side, it's more,
https://karpathy.ai/lexicap/0012-large.html#00:59:51.040
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
so global warming is a more encroaching problem.
https://karpathy.ai/lexicap/0012-large.html#00:59:53.560
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Nuclear weapons have been here.
https://karpathy.ai/lexicap/0012-large.html#01:00:00.840
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
It's an obvious problem that's just been sitting there.
https://karpathy.ai/lexicap/0012-large.html#01:00:03.280
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So how do you think about,
https://karpathy.ai/lexicap/0012-large.html#01:00:05.720
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
what is the mechanism design there
https://karpathy.ai/lexicap/0012-large.html#01:00:07.480
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
that just made everything seem stable?
https://karpathy.ai/lexicap/0012-large.html#01:00:09.240
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And are you still extremely worried?
https://karpathy.ai/lexicap/0012-large.html#01:00:12.280
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
I am still extremely worried.
https://karpathy.ai/lexicap/0012-large.html#01:00:14.800
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So you probably know the simple game theory of mad.
https://karpathy.ai/lexicap/0012-large.html#01:00:16.640
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So this was a mutually assured destruction
https://karpathy.ai/lexicap/0012-large.html#01:00:20.040
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
and it doesn't require any computation with small matrices.
https://karpathy.ai/lexicap/0012-large.html#01:00:23.760
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
You can actually convince yourself
https://karpathy.ai/lexicap/0012-large.html#01:00:27.360
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
that the game is such that nobody wants to initiate.
https://karpathy.ai/lexicap/0012-large.html#01:00:28.600
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Yeah, that's a very coarse grained analysis.
https://karpathy.ai/lexicap/0012-large.html#01:00:31.480
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And it really works in a situational way.
https://karpathy.ai/lexicap/0012-large.html#01:00:34.600
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
You have two superpowers or small number of superpowers.
https://karpathy.ai/lexicap/0012-large.html#01:00:36.880
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Now things are very different.
https://karpathy.ai/lexicap/0012-large.html#01:00:40.400
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
You have a smaller nuke.
https://karpathy.ai/lexicap/0012-large.html#01:00:41.960
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So the threshold of initiating is smaller
https://karpathy.ai/lexicap/0012-large.html#01:00:43.080
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
and you have smaller countries and non nation actors
https://karpathy.ai/lexicap/0012-large.html#01:00:47.240
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
who may get a nuke and so on.
https://karpathy.ai/lexicap/0012-large.html#01:00:51.520
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
So I think it's riskier now than it was maybe ever before.
https://karpathy.ai/lexicap/0012-large.html#01:00:53.760
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
And what idea, application of AI,
https://karpathy.ai/lexicap/0012-large.html#01:00:58.320
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
you've talked about a little bit,
https://karpathy.ai/lexicap/0012-large.html#01:01:03.640
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
but what is the most exciting to you right now?
https://karpathy.ai/lexicap/0012-large.html#01:01:04.640
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
I mean, you're here at NIPS, NeurIPS.
https://karpathy.ai/lexicap/0012-large.html#01:01:07.560
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
Now you have a few excellent pieces of work,
https://karpathy.ai/lexicap/0012-large.html#01:01:10.160
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
but what are you thinking into the future
https://karpathy.ai/lexicap/0012-large.html#01:01:14.920
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
with several companies you're doing?
https://karpathy.ai/lexicap/0012-large.html#01:01:16.680
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
What's the most exciting thing or one of the exciting things?
https://karpathy.ai/lexicap/0012-large.html#01:01:17.840
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
The number one thing for me right now
https://karpathy.ai/lexicap/0012-large.html#01:01:21.120
Tuomas Sandholm: Poker and Game Theory | Lex Fridman Podcast #12
is coming up with these scalable techniques
https://karpathy.ai/lexicap/0012-large.html#01:01:23.160