episode
stringlengths
45
100
text
stringlengths
1
528
timestamp_link
stringlengths
56
56
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and that it's much more scalable,
https://karpathy.ai/lexicap/0017-large.html#00:09:21.440
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
that you absolutely want that thing
https://karpathy.ai/lexicap/0017-large.html#00:09:23.880
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
to go read the whole scientific literature
https://karpathy.ai/lexicap/0017-large.html#00:09:25.880
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and think about how to create cures for all the diseases,
https://karpathy.ai/lexicap/0017-large.html#00:09:27.640
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
right?
https://karpathy.ai/lexicap/0017-large.html#00:09:29.820
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
You want it to think about how to go
https://karpathy.ai/lexicap/0017-large.html#00:09:30.660
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and build technologies to help us create material abundance
https://karpathy.ai/lexicap/0017-large.html#00:09:31.480
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and to figure out societal problems
https://karpathy.ai/lexicap/0017-large.html#00:09:34.500
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
that we have trouble with.
https://karpathy.ai/lexicap/0017-large.html#00:09:37.320
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
Like how are we supposed to clean up the environment?
https://karpathy.ai/lexicap/0017-large.html#00:09:38.160
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And maybe you want this to go and invent
https://karpathy.ai/lexicap/0017-large.html#00:09:40.000
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
a bunch of little robots that will go out
https://karpathy.ai/lexicap/0017-large.html#00:09:42.840
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and be biodegradable and turn ocean debris
https://karpathy.ai/lexicap/0017-large.html#00:09:44.120
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
into harmless molecules.
https://karpathy.ai/lexicap/0017-large.html#00:09:47.280
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And I think that that positive side
https://karpathy.ai/lexicap/0017-large.html#00:09:49.660
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
is something that I think people miss
https://karpathy.ai/lexicap/0017-large.html#00:09:54.040
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
sometimes when thinking about what an AGI will be like.
https://karpathy.ai/lexicap/0017-large.html#00:09:55.720
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And so I think that if you have a system
https://karpathy.ai/lexicap/0017-large.html#00:09:58.160
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
that's capable of all of that,
https://karpathy.ai/lexicap/0017-large.html#00:10:00.280
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
you absolutely want its advice about how do I make sure
https://karpathy.ai/lexicap/0017-large.html#00:10:01.640
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
that we're using your capabilities
https://karpathy.ai/lexicap/0017-large.html#00:10:03.960
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
in a positive way for humanity.
https://karpathy.ai/lexicap/0017-large.html#00:10:07.600
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
So what do you think about that psychology
https://karpathy.ai/lexicap/0017-large.html#00:10:09.220
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
that looks at all the different possible trajectories
https://karpathy.ai/lexicap/0017-large.html#00:10:11.440
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
of an AGI system, many of which,
https://karpathy.ai/lexicap/0017-large.html#00:10:14.800
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
perhaps the majority of which are positive,
https://karpathy.ai/lexicap/0017-large.html#00:10:17.520
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and nevertheless focuses on the negative trajectories?
https://karpathy.ai/lexicap/0017-large.html#00:10:19.960
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
I mean, you get to interact with folks,
https://karpathy.ai/lexicap/0017-large.html#00:10:23.340
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
you get to think about this, maybe within yourself as well.
https://karpathy.ai/lexicap/0017-large.html#00:10:24.720
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
You look at Sam Harris and so on.
https://karpathy.ai/lexicap/0017-large.html#00:10:28.860
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
It seems to be, sorry to put it this way,
https://karpathy.ai/lexicap/0017-large.html#00:10:30.560
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
but almost more fun to think about
https://karpathy.ai/lexicap/0017-large.html#00:10:32.760
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
the negative possibilities.
https://karpathy.ai/lexicap/0017-large.html#00:10:34.560
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
Whatever that's deep in our psychology,
https://karpathy.ai/lexicap/0017-large.html#00:10:36.780
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
what do you think about that?
https://karpathy.ai/lexicap/0017-large.html#00:10:39.560
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And how do we deal with it?
https://karpathy.ai/lexicap/0017-large.html#00:10:40.840
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
Because we want AI to help us.
https://karpathy.ai/lexicap/0017-large.html#00:10:41.920
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
So I think there's kind of two problems
https://karpathy.ai/lexicap/0017-large.html#00:10:44.400
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
entailed in that question.
https://karpathy.ai/lexicap/0017-large.html#00:10:47.360
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
The first is more of the question of
https://karpathy.ai/lexicap/0017-large.html#00:10:49.960
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
how can you even picture what a world
https://karpathy.ai/lexicap/0017-large.html#00:10:52.360
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
with a new technology will be like?
https://karpathy.ai/lexicap/0017-large.html#00:10:54.620
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
Now imagine we're in 1950,
https://karpathy.ai/lexicap/0017-large.html#00:10:56.600
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and I'm trying to describe Uber to someone.
https://karpathy.ai/lexicap/0017-large.html#00:10:57.840
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
Apps and the internet.
https://karpathy.ai/lexicap/0017-large.html#00:11:02.840
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
Yeah, I mean, that's going to be extremely complicated.
https://karpathy.ai/lexicap/0017-large.html#00:11:05.340
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
But it's imaginable.
https://karpathy.ai/lexicap/0017-large.html#00:11:08.920
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
It's imaginable, right?
https://karpathy.ai/lexicap/0017-large.html#00:11:10.160
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And now imagine being in 1950 and predicting Uber, right?
https://karpathy.ai/lexicap/0017-large.html#00:11:11.880
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And you need to describe the internet,
https://karpathy.ai/lexicap/0017-large.html#00:11:15.280
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
you need to describe GPS,
https://karpathy.ai/lexicap/0017-large.html#00:11:17.680
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
you need to describe the fact that
https://karpathy.ai/lexicap/0017-large.html#00:11:18.720
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
everyone's going to have this phone in their pocket.
https://karpathy.ai/lexicap/0017-large.html#00:11:20.480
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And so I think that just the first truth
https://karpathy.ai/lexicap/0017-large.html#00:11:23.920
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
is that it is hard to picture
https://karpathy.ai/lexicap/0017-large.html#00:11:26.160
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
how a transformative technology will play out in the world.
https://karpathy.ai/lexicap/0017-large.html#00:11:28.040
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
We've seen that before with technologies
https://karpathy.ai/lexicap/0017-large.html#00:11:31.160
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
that are far less transformative than AGI will be.
https://karpathy.ai/lexicap/0017-large.html#00:11:32.760
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And so I think that one piece is that
https://karpathy.ai/lexicap/0017-large.html#00:11:35.560
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
it's just even hard to imagine
https://karpathy.ai/lexicap/0017-large.html#00:11:37.780
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and to really put yourself in a world
https://karpathy.ai/lexicap/0017-large.html#00:11:39.560
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
where you can predict what that positive vision
https://karpathy.ai/lexicap/0017-large.html#00:11:41.640
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
would be like.
https://karpathy.ai/lexicap/0017-large.html#00:11:44.640
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And I think the second thing is that
https://karpathy.ai/lexicap/0017-large.html#00:11:46.920
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
I think it is always easier to support the negative side
https://karpathy.ai/lexicap/0017-large.html#00:11:49.520
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
than the positive side.
https://karpathy.ai/lexicap/0017-large.html#00:11:54.280
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
It's always easier to destroy than create.
https://karpathy.ai/lexicap/0017-large.html#00:11:55.120
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And less in a physical sense
https://karpathy.ai/lexicap/0017-large.html#00:11:58.160
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and more just in an intellectual sense, right?
https://karpathy.ai/lexicap/0017-large.html#00:12:00.760
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
Because I think that with creating something,
https://karpathy.ai/lexicap/0017-large.html#00:12:03.080
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
you need to just get a bunch of things right.
https://karpathy.ai/lexicap/0017-large.html#00:12:05.680
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And to destroy, you just need to get one thing wrong.
https://karpathy.ai/lexicap/0017-large.html#00:12:07.400
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And so I think that what that means
https://karpathy.ai/lexicap/0017-large.html#00:12:10.280
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
is that I think a lot of people's thinking dead ends
https://karpathy.ai/lexicap/0017-large.html#00:12:12.040
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
as soon as they see the negative story.
https://karpathy.ai/lexicap/0017-large.html#00:12:14.240
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
But that being said, I actually have some hope, right?
https://karpathy.ai/lexicap/0017-large.html#00:12:16.920
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
I think that the positive vision
https://karpathy.ai/lexicap/0017-large.html#00:12:20.360
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
is something that I think can be,
https://karpathy.ai/lexicap/0017-large.html#00:12:23.160
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
is something that we can talk about.
https://karpathy.ai/lexicap/0017-large.html#00:12:26.000
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And I think that just simply saying this fact of,
https://karpathy.ai/lexicap/0017-large.html#00:12:27.560
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
yeah, there's positive, there's negatives,
https://karpathy.ai/lexicap/0017-large.html#00:12:30.240
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
everyone likes to dwell on the negative.
https://karpathy.ai/lexicap/0017-large.html#00:12:32.000
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
People actually respond well to that message and say,
https://karpathy.ai/lexicap/0017-large.html#00:12:33.600
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
huh, you're right, there's a part of this
https://karpathy.ai/lexicap/0017-large.html#00:12:35.400
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
that we're not talking about, not thinking about.
https://karpathy.ai/lexicap/0017-large.html#00:12:37.040
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
And that's actually something that's I think really
https://karpathy.ai/lexicap/0017-large.html#00:12:39.640
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
been a key part of how we think about AGI at OpenAI.
https://karpathy.ai/lexicap/0017-large.html#00:12:42.280
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
You can kind of look at it as like, okay,
https://karpathy.ai/lexicap/0017-large.html#00:12:46.640
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
OpenAI talks about the fact that there are risks
https://karpathy.ai/lexicap/0017-large.html#00:12:48.640
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
and yet they're trying to build this system.
https://karpathy.ai/lexicap/0017-large.html#00:12:51.040
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
How do you square those two facts?
https://karpathy.ai/lexicap/0017-large.html#00:12:53.720
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
So do you share the intuition that some people have,
https://karpathy.ai/lexicap/0017-large.html#00:12:56.120
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
I mean from Sam Harris to even Elon Musk himself,
https://karpathy.ai/lexicap/0017-large.html#00:12:59.160
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
that it's tricky as you develop AGI
https://karpathy.ai/lexicap/0017-large.html#00:13:02.720
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
to keep it from slipping into the existential threats,
https://karpathy.ai/lexicap/0017-large.html#00:13:06.640
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
into the negative?
https://karpathy.ai/lexicap/0017-large.html#00:13:10.440
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
What's your intuition about how hard is it
https://karpathy.ai/lexicap/0017-large.html#00:13:11.800
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
to keep AI development on the positive track?
https://karpathy.ai/lexicap/0017-large.html#00:13:14.840
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
What's your intuition there?
https://karpathy.ai/lexicap/0017-large.html#00:13:19.680
Greg Brockman: OpenAI and AGI | Lex Fridman Podcast #17
To answer that question, you can really look
https://karpathy.ai/lexicap/0017-large.html#00:13:20.760