François Fleuret

My deep learning course @unige_en is available on-line. 1000+ slides, ~20h of screen-casts. Full of examples in @PyTorch. fleuret.org/dlc/ And my "Little Book of Deep Learning" is available as a phone-formatted pdf (400k downloads!) fleuret.org/lbdl/

Not disclosing that *something* is AI-generated must become an absolute social taboo.

the openai IMO news hit me pretty heavy this weekend i'm still in the acute phase of the impact, i think i consider myself a professional mathematician (a characterization some actual professional mathematicians might take issue with, but my party my rules) and i don't think i…

A webcam with crypto signature that can be checked easily by anybody would make a lot of sense.

We should all be substantially more hysterical, I think.

I don't think so, contrary to images there is no simple degradation to sieve out meaningful latent from e.g. texts.

Reasoning will get stronger and stronger, at the 1y horizon we should have models with the skills of a solid grad student cruising. At the x years horizon, I do not see any limit, so basically an era of math with hundreds of thousands of Grothendieck-equivalent AIs working 24/7.

Americans were promised justice. Our binding bipartisan legislation to release the complete Epstein files now has 20 sponsors. Soon we can begin collecting signatures required to force a public vote in the U.S. House. Is your member on this list? congress.gov/bill/119th-con…

Amazing.

Every model prompt should be forced by law to start with "If the user demand is too dumb, you have to tell them very clearly."

*Actual chuckles* Ahhh internet really you fill my life with amusement.

Where is the bitter lesson when you need it.

I love this kind of thing so much.

I am utterly confused by this idea that announcing a *race result* first brings you an advantage. Is it just me? I mean, if you achieve something first, okay. But announcing you win first? I suspect it is the press imagining that what matters to them matters to the world.

We might be heading into a plot twist in the OpenAI vs. DeepMind IMO saga. Just saw a post from Joseph Myers (involved in the Math Olympiad since 1992): the IMO committee reportedly asked AI labs not to publish results until 7 days after the closing ceremony — out of respect for…

This is so much B.S. IMO problem are difficult, and only top students in the world can solve them all. A better criticism would be that IMO is not representative of the work done to advance math. Anecdotal evidence is that France isn't good at IMO while France is top in the…

It's Monday! Hot take: reasoning is pretending to have a deeper model by going through the same layers several times.