Ne Zha

[Ne Zha @NeZhakyh]

soul selects her own society, and then shuts the door

https://t.co/seVgWunyPh

[Ne Zha @NeZhakyh]

@suchenzang You don’t understand. I was just doing human collaboration.

[Ne Zha @NeZhakyh]

@nihilunbounded 🐍

[Ne Zha @NeZhakyh]

@nihilunbounded can’t drop fast enough

[Linus Mixson @LinusMixson]

1. I seem to recall Anthropic scooping Kevin Buzzard's "personal mathematical collaboration," with no attempt to include him at all, not one week ago.
2. > spend months doing quirky lower-case drops of zero-effort proofs from your lab that obviate years of some researcher's diligent labor > have your friend's labor obviated by a competing lab > post weird weepy lowercase personal anecdote about how sad it is.
3. Is he talking about reassuring Astra, Astra the model, like this is a significant point in his ethical favor? What the fuck? What is Amanda doing over there?

[Joel Fish @joelwatsonfish]

But could we have a conference on keeping mathematicians safe from AI?

[Ne Zha @NeZhakyh]

@nihilbounded with gabbers post about jacobian and this about soficity… suddenly reality seems different.

[Ne Zha @NeZhakyh]

@yiannissak also look at Thoms post on mathstodon on the nonsofic group. tldr: he had been discussing that approach with gpt and never got straight amswer if his conversations were used. He found it eerie that there were other more promising approaches that model didnt take

[Alvaro Lozano-Robledo @mathandcobb]

Btw Andreas Thom is now also wondering outloud whether OpenAI used his personal user data to train the model to find a non-sofic group...
https://t.co/6uOUNWCUHN

[doomslide @doomslide]

@__alpoge__ This wall of text cannot hide the small, petty mind behind it. You're a collosal disgrace to the mathematics profession Levent. I hope it was worth it.

[Ne Zha @NeZhakyh]

that’s it. Not posting papers on arxiv anymore.

[Genius Ge @Soledad7000]

Hey @__alpoge__ this new Borisov-Gabber-Vasiu preprint suggests that an https://t.co/He83aHIar8 paper explaining the counterexample to the Jacobian conjecture drew from an earlier version of their work. Care to explain how you obtained the counterexample? https://t.co/A3YilnQABW

[Ne Zha @NeZhakyh]

@tonyyueyu when people say they don’t use AI they mean they don’t use it to discuss ideas or generate proofs and such.

[Lior Pachter @lpachter]

Mathematics has now fully morphed into biology:

Project costs millions $ ✅
Put out paper that nobody has read ✅
Massive paper supplement nobody will read ✅
Advertise with pretty art ✅
Vicious authorship fight ✅
Say you're curing cancer ✅

[doomslide @doomslide]

@EAgony580 I wasn't making an analogy. LLMs are mimicry (pretrain) + hill climbing (RL). That's a factual claim about what the technology is. The hill climbing part can be used to sharpen the pretrained distribution but it's still executing the same text patterns from the data.

[Ne Zha @NeZhakyh]

@torgeirlysen 1000year butlerian jihaad now.

[Ne Zha @NeZhakyh]

how does he consistently have the most retarded takes. Did working with bourgain stunt his development?

[Alex Kontorovich @AlexKontorovich]

My answer: people like me will get to work on finding a shorter (or perhaps longer, if need be!), human-understandable explanation. If there’s Lean code, then I’d start iterating with an AI, tell it all the parts I understand so far, and start trying to figure out the first thing I don’t already know.

In fact, if I really wanted to stop my research for a while, I would start doing this right now with FLT, and learn more of how the proof really goes!

Having a Lean formalization opens up a whole new world of possibilities for human understanding!

[netsky remix @iron_redux]

PUT HIM IN A WORK CAMP Десять лет без права переписки

[Ne Zha @NeZhakyh]

@suchenzang she hasn’t stopped yet. I wonder of she has an agent doing it for her.

[dzackgarza @dzackgarza]

And it seems clear to me that what used to be called hallucination has been RL'd into pathological, borderline sociopathic lying and misdirection, with the world as its eval. No amount of expert agreement or certificates ameliorates the fact that that it's trained to fool people

[Ne Zha @NeZhakyh]

@suchenzang ah that right. Tbh he is famous for his earlier work with Vicol so he has a lot of glory to go around. In fact writing it with apologe gets him less credit/prestige.