@andrewgwils Industry does do quite a lot of fundamental research. The frontiers of fundamental research (in ML) and practice have merged, but only very few people can see and understand this. Eg, Imho, pretraining is the frontier of the most fundamental research in machine learning .

@Ar_Douillard solid paper. 14 pages reference 🤣

@sfrei_ @GoogleDeepMind Welcome! Glad to have you joined us!

Very cool! Congrats and thanks for trusting our codebase!+ @peterjliu (creator of nanodo), @ARomanNovak ,@_katieeverett (pre-)lead of the codebase.

It’s extremely exciting to see the frontiers of science and practice merge, yet equally disheartening to see when they become frontier of geopolitics. What wisdoms do AI offer to human?

@kellerjordan0 @adamlerer @mobav0 Congrats and thanks for your inspiring work to the community!

@spteotia Possible if I live long enough

@spteotia L5 is certainly asi level ; l6 may be solving several of them

@BrandoHablando Sth likes several good results that each can be published in a top tier 2 math journal (eg transaction of American mathematical society), top journal in a subfield (eg analysis and pde); ~ a good phd thesis in a top 20 phd program

@dmdohan any thoughts about how the difficulty levels (low, medium, high) distributed across the 25% of problems that have been solved?

@cloneofsimo fyi, this particular problem was in the high difficulty category of the dataset; there are low/medium difficulty problems, presumably easier; see appendix A.1

@jasondeanlee not research problems; sth more like (hard) GTM exercises

“勿以恶小而为之，勿以善小而不为。惟贤惟德，能服于人” ("Do not do evil just because it is small; do not ignore doing good just because it is small. Only by being virtuous and moral can one be respected by others.") learned this in middle school, never forgot, passed it to my kids

@_arohan_ You will be missed and thanks for your many contributions.

@hyhieu226 Why not coscale it with #params?

@giffmana @__kolesnikov__ @XiaohuaZhai This should be 0. lol . Sad to c u leave , congrats oai

@Grad62304977 Xlable: step

RT @alemi: If you miss the NYTimes needle, especially one that is statistically uniform (, you can use this page: h…

Lgtm