"Where I agree and disagree with Eliezer" by Paul Christiano
LessWrong (Curated & Popular)
00:00
The Disjunctiveness of Alignment Hopes
I think that the last five years of progress in language modeling have provided significant evidence that training AI to imitate human thought may be economically competitive. It seems plausible that AI systems will learn much of how to think by predicting humans, even if human language is a uselessly shallow shadow of human thought. I can't tell if Ellie has a should have lost base points here, but I suspect he would have. Most importantly, it seems like AI systems have huge structural advantages, like their high speed and low cost. That suggests they will have a transformative impact on the world, and obsolete human contributions to alignment.
Play episode from 26:04
Transcript
