Christian Szegedy (@ChrSzegedy) / X

Christian Szegedy
@ChrSzegedy
22h
Re: community note In order to avoid PR backlash, mathematicians might want to start The Association of Sane Mathematics
Ellie Sleightholm
@elsleightholm
Oct 7
Important statement from the Association for Human Mathematics ahmath.org/statements
5
Christian Szegedy
@ChrSzegedy
22h
Aligning AI is hopeless. Let's align the user.
35
Christian Szegedy
@ChrSzegedy
Oct 8
I just chatted with my brother Balazs about OpenAI's papers. He was super pumped about some of the results. One thing, of course, is the matrix multiplication paper. He used to work on this 20 years ago and said that he located an ingenious-looking computational trick in the new
9
Christian Szegedy
@ChrSzegedy
Oct 8
Up until last year, all these results would be papers separated by years, waiting for peer review for months, communicated in arxiv papers first, published after a year in some journal. Now, all that research using AI and formal tools can be compressed into hours.
Aurel Prosz
@aurel_pr
Oct 8
I've just made a live tracker for the community effort for optimizing the "Integer multiplication below n log n" mathematical problem, so everybody can see the progress on it: beyond-n-log-n.netlify.app
6
Christian Szegedy
@ChrSzegedy
Oct 8
That's why I believe strongly that formal verification will be a critical component of being able to use superhuman AI. Formal verification is a cheat code by which a really dumb program can check an output produced by a superintelligent entity.
Frank
@CouldBeWrongTho
Oct 8
Replying to @ChrSzegedy @oydeis and @Quasilocal
An expected outcome of increasingly intelligent entities is that people using them increasingly dont understand what they are saying. It already happens between humans. It will happen more and more with AI.
13