Menu
Sign In Search Podcasts Charts People & Topics Add Podcast API Pricing

Eliezer Yudkowsky

๐Ÿ‘ค Speaker
See mentions of this person in podcasts
1713 total appearances

Appearances Over Time

Podcast Appearances

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

Like...

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

The inside of a brick is not like a hollow shell containing only a surface.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

There's an inside of the brick.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

If you put it into an x-ray machine, you can see the inside of the brick.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

And, you know, just because we cannot understand what's going on inside GPT does not mean that it is not there.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

A blank map does not correspond to a blank territory.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

I think it is like predictable...

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

with near certainty that if we knew what was going on inside GPT, or let's say GPT-3 or even like GPT-2 to take one of the systems that like has actually been open sourced by this point, if I recall correctly,

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

If we knew it was actually going on there, there is no doubt in my mind that there are some things it's doing that are not exactly what a human does.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

If you train a thing that is not architected like a human to

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

predict the next output that anybody on the internet would make.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

This does not get you this agglomeration of all the people on the internet that rotates the person you're looking for into place and then simulates the internal processes of that person one-to-one.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

It is to some degree an alien actress.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

It cannot possibly just be a bunch of different people in there, exactly like the people.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

But how much of it is by gradient descent getting optimized to perform similar thoughts as humans think in order to predict human outputs versus being optimized to...

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

carefully consider how to play a role, like how humans work, predict the actress, the predictor, that in a different way than humans do.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

Well, you know, that's the kind of question that with like 30 years of work by half the planet's physicists, we can maybe start to answer.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

You think so?

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

Well, there's a difference.

Lex Fridman Podcast
#368 โ€“ Eliezer Yudkowsky: Dangers of AI and the End of Human Civilization

So I think even GPT-2 probably has very stupid fragments of alien actress in it.