Poll: Can generative AI solve computer science's greatest unsolved "Does P = NP" problem? #compsci Please boost for reach.
yes
6.3%
noop
93.7%
Poll ended at .
@nixCraft don't need an AI to know that P = 1??? /s
@nixCraft Lol, it can't even follow the rules of chess.
@nixCraft I don’t even understand the basis of the question. What makes you think it could?
@nixCraft unless you consider a generous hallucination! Then you can "prove" anything.
@nixCraft
Today's LLMs are designed to string words together plausibly based on previous observances. They can't do math because they understand relative word frequencies but not actual relationships between words. If an LLM's data inputs happened upon somebody's mathematical proof, it might be able to talk about it convincingly, but it would still get the details wrong.
@nixCraft at the moment, it fails to answer pretty much any question of interest I throw at it.
@nixCraft I think we should stop calling "Large Language Models", AI.
Is there any problem that could be solved by LLM?
@nixCraft
But can generative AI do so in np time?
@nixCraft How about you first finish building that big ol’ #NTM and we’ll see how close we can get with the three-taper from the #nixCraft basement, right after?! 😅💨

@nixCraft
Yes -- given an arbitrary number of LLM generators and a sufficiently long length of time, a generative 'AI' may return a solution which is correct and contains a suitably detailed proof. That said, the heat death of the universe may not be sufficient time.

Of course, nobody asks who's going to feed the infinite monkeys...

@nixCraft I'm still not sure if "noop" is supposed to be a long "nope" or a machine code "no-op", but both work here.
@nixCraft I would say possibly... eventually :-)