Eventually I cut out all the "Uh"s from the recording, almost the same effect as a 2x speedup :-) (additionally had to cut just a few sentences at the end).
Recorded a video presentation that is now required in addition to the conference submission... but it's one hour long! Should I speed it up 2x before uploading? :-)
@proofchecker@mayankjha1089 No, they did not formalize it before publishing ("extent possible" meant they could not wait for another week). And they published it with much hype, which counts more than arXiv.
@iliallusion Кажется, "папа у Васи силен в математике" написано не в связи с ИИ. Для младших и средних классов это вопрос ресурсов и воспитания, а не ИИ. Для старших да, надо что-то менять.
@mayankjha1089 I am not saying humans don't publish flawed proofs: we do. The question is how frequent and how important. Humans take this risk and publish. Machines don't care about their reputation, human relations, and food, - their publishing attempts are riskless.
@mayankjha1089 It means something very imporant: the authors put their reputation at risk regarding, say, publishing flawed proofs. This is exactly what was missing in OpenAI's write-ups: a human being putting their reputation at stake.
@StefanSzeider Looks surprising to me. I would expect detailed instructions concerning some harness rather than reminders that the the order of quantifiers matters. I think even publicly avaliable models nowadays distinguish a proof from a non-proof on that scale.
@EliBenSasson@avihu28 I do not think the quantum threat is real. These toys are sooo much expensive that I would compare it to a threat from LHC or Space Shuttle 30 years ago. If the US govt wants to hack you personally... well, may happen.
@diveshaggarwal It's unclear to exactly the same extent it's unclear where we'll be living irrespectively of tcs. Other than that, there're plenties of directions, more than ever; room for everyone. Our community needs to explain it to the administration and to the prospective 1st year students