Gary Marcus

Rank 17 of 47

|

Score 63

·

9h

@Vector76 i don’t think Danielle Steele is slop per se, and novel length poses problems of coherence and originality that LLMs aren’t well positioned to address.

·

9h

@NathanpmYoung before gpt-5.X or GPT-6 comes out. the claim is just that the current version can’t do it.

·

9h

ok, time for a new bet: I bet that GPT-5 can’t write a romance novel (without extensive plagiarism) that some reasonable panel of judges finds readable enough to make it through to the end.

·

9h

new theory: what Ilya saw was that … AGI porn was not in fact going to be all that revolutionary

·

10h

@shi_weiyan not super diverse and very problematic, as we move out of distribution.

·

11h

@shi_weiyan way better than earlier though some still not actually riding, and of course “horse rides astronaut” is a prompt i have been writing about for years”, so may have been a focus of training.

Could you try my more-recent, less discussed “draw a horse riding an astronaut wearing an

·

12h

@OlesSeymour @deredleritt3r its preclinical, so we don’t know yet.

·

12h

@chatgpt21 they will certainly try. pure LLMs won’t get us there then, though.

i think also that there is a difference between predicting the logic of why the economics would fail (which seems to be basically right) vs guessing the date on which investors would be correct (a matter of

·

12h

@Trachurus would actually be interesting to real numbers on that

·

12h

@JoannaStern Then again, have sext ≠ have sex?

It all depends on what the meaning of “is” is?