레커:

I projected it would continue at this pace. I figured we would pass the Turing test within 30 years by 2029.

Uh, Stanford felt that was very alarming, and so they held an international conference, and AI experts came from all over the world. Um, and, uh, they felt I was very over optimistic. Uh, yeah.


사회자:

Maybe show of hands today, who thinks we're gonna beat 2029?


레커:

So this, um, they felt it would take a hundred years.

So they felt it would happen, but they felt it would take a hundred years. Um, now, now my, I'm still saying 2029, and it turns out to be pessimistic.

A lot of people are saying next year, two years, some people think it's already happened.

Uh, the Turing test is actually not very well defined. Turing wrote an essay about it.

There's maybe a few, uh, lines that describe what the Turing test is, and people disagree on it.

Uh, so I figured people would say we're passing a Turing test, but it wouldn't be real until maybe most people felt that we're passing a Turing test.

So I think that'll actually start next year. And by 2029 will be the kind of the last predictions.

Um, it turns out that our ability to actually create something that's super human is, is around the same speed because already, if you look at a large language model, there's certain things that can't do that humans can do. But there's lots of things it can do that, that humans can't do.

(후략)


자기 예측은 똑같은데 과거랑 지금이랑 취급이 다르다는 얘기는 불과 며칠 전에 특갤 념글로도 갔었던 주제인데 (조 로건하고 한 팟캐스트에서도 비슷한 얘길 했거든)

빌게이츠는 기사 정리한 애 있으니까 생략

새로운 떡밥이면 특붕이들 행복회로 절망회로 돌리는 거 구경하는 재미라도 있는데

옛날하고 다른 말을 한 게 없는데 어거지로 끄집어내서 회로 돌리는 건 재미없고 지루함