none

[none @nihilsuncle]

@nihilunbounded @ElliotGlazer oh boy...

[none @nihilsuncle]

@davikrehalt @KnightShore also people could take some time and try to actually write a nice paper. my complaint is specifically towards non-experts quickly writing/modifying a correct but meh writeup of an LLM solution

[none @nihilsuncle]

"Mathematicians", let's cut it out with "the author independently verified all... and takes full responsibility for the content of the manuscript." At this stage, ChatGPT can verify itself better than you can. If you don't contribute an actual major idea, leave your name off.

[none @nihilsuncle]

@davikrehalt @KnightShore If/when systems get powerful enough, then anyone can just ask whatever they want to know anyways. For now, credit can still matter and this just adds a lot of noise.

[none @nihilsuncle]

@PI010101 @littmath In the human era, the case where general suggestions were given was a borderline situation where people followed their own ethics and conscience.

[none @nihilsuncle]

@PI010101 @littmath I can live with this later case. Personally I usually gave away a lot of things to people when I didn't want to do the actual work rather than try to stuff my name on it, but OK.

[none @nihilsuncle]

@littmath @Soledad7000 I don't find a lot of these "expositions" to be J.P. Serre level.

[none @nihilsuncle]

@MCktm5 @shallit43 I don't claim it's perfect. But if you use agents in the right way, in most subfields I think its accuracy is comparable to a professional mathematician and certainly a non-expert trying to YOLO a problem list.

[none @nihilsuncle]

@KnightShore Definitely not the case with GPT-6 Astra if you are a sufficiently good mathematician. Maybe (probably) true in the limiting case. Probably, in that case most papers are not worth putting out (note: I don't claim to have a good path forward).

[none @nihilsuncle]

@dsavitt @littmath I don't think you should write papers that have none of your ideas.

[none @nihilsuncle]

@TerreMath @littmath The situation is sad for me but it is what it is. I don't believe people doing ~nothing and trying to slip in papers before the situation has clarified will look good.

[none @nihilsuncle]

@bbzz1067029 This is hard because there is an economic question. Cannot be solved by $5000 of API credits, $50,000?

[none @nihilsuncle]

@AB622565132484 Fine with me. Or just have people sign it X, conduit of ChatGPT/Claude. Authorship means ideas, not referee work. If some mathematicians have no ideas not in the system too bad.

[none @nihilsuncle]

@Soledad7000 if you use smart agent work it's pretty careful. Strong experts in very technical branches can still catch things but those are not usually the ones producing this stuff.

[none @nihilsuncle]

@torgeirlysen Is this a P(doom) type statement? Then, sure, but these meetings ain't it.

[none @nihilsuncle]

One predictable (but dispiriting) trend in the academic response to math x AI is the consolidation of power. Meetings like this https://t.co/z2KPs4iegq involving closed circles of (American) mathematical ``leaders" who aim to set guidelines for others. (1/1)

[none @nihilsuncle]

@pilkself I prefer to poast.

[none @nihilsuncle]

@FrankT56966988 They could invite people of "lower status" or even diversify which "high class" people contribute. It appears to be mostly the same group of 10 or so who may be more willing to pontificate...

[none @nihilsuncle]

In the worst case scenario, they have nice checks and a lot of free time to garden on their hands (3/3).

[none @nihilsuncle]

Yet I've yet to say any of these ``leaders" say something exceptionally brilliant about the situation. There may, in fact, be nothing exceptionally brilliant to say. But surely there is nothing particularly at stake for them. It is unlikely any of the departments will close. (2/2