XY Han

Assistant Professor @ChicagoBooth | Most Cited Paper: “Neural Collapse in Deep Net Training” | BSE @Princeton, PhD @Cornell, MS/PostDoc @Stanford

[zeta @zeta_globin]

there has never been anything I've ever wanted for a birthday past the age of 8 besides being completely alone and having absolutely nothing scheduled so I can sleep, get coffee, clean my place, go for walk, call my parents since really it's their day, and speak to no one else

[Andrew Gordon Wilson @andrewgwils]

Perhaps I'm an outlier, but generally the value I derive from art is not from its backstory. I love a Bach fugue not because he was suffering, content, had many children, or whatever else, but because it's an extraordinary composition. I'd feel the same about AI generated art.

[Weijie Su @weijie444]

Personal update: I've joined OpenAI while on leave from Wharton. After a decade away, glad to be back in the Bay Area and train AI models here!

One more thing, I've been promoted to full professor, a decade-long journey made possible by many, especially my students.

[XY Han @XYHan_]

@weijie444 I heard Emmanuel had a former student who got the COPSS prize this year too. 😉

[Weijie Su @weijie444]

Wow, my former PhD advisor got the Shaw Prize!

[XY Han @XYHan_]

Best one I’ve seen so far.

[Ash Jogalekar @curiouswavefn]

Freeman Dyson: “A scientist is a person who comes up with new ideas. An engineer is a person who makes something work with as few new ideas as possible.”

[Ash Jogalekar @curiouswavefn]

Freeman Dyson: “A scientist is a person who comes up with new ideas. An engineer is a person who makes something work with as few new ideas as possible.”

[Lior Pachter @lpachter]

This person has published 71 papers in 143 days so far in 2026. That is, 2 days per paper (source: Google Scholar).

It's truly amazing. To see someone proud of this.

[David Sun @arcticinstincts]

Bryan Johnson rediscovered Asian Aunties from first principles

[XY Han @XYHan_]

@alexolegimas The cafe in the Hilton in South Loop (next to Grant Park) actually sells deep dish by the slice. Not sure if it’s good, but they do it… 😅

[Arthur Spirling @arthur_spirling]

their locations were selected to facilitate the teaching of hard margin SVM. few know this

[Burny - Effective Curiosity @burny_tech]

"When Ronald Graham, a concerned friend and fellow mathematician, bet Erdos $500 that he couldn’t stay off amphetamines for a month, Erdos accepted and easily won the challenge. When the 30 days was up, Erdos said to Graham:

“You’ve showed me I’m not an addict. But I didn’t get any work done. I’d get up in the morning and stare at a blank piece of paper. I’d have no ideas, just like an ordinary person.“

“You’ve set mathematics back a month.”

Erdos resumed taking amphetamines and did so for every day of his life until his death 17 years later."

[Burny - Effective Curiosity @burny_tech]

"When Ronald Graham, a concerned friend and fellow mathematician, bet Erdos $500 that he couldn’t stay off amphetamines for a month, Erdos accepted and easily won the challenge. When the 30 days was up, Erdos said to Graham:

“You’ve showed me I’m not an addict. But I didn’t get any work done. I’d get up in the morning and stare at a blank piece of paper. I’d have no ideas, just like an ordinary person.“

“You’ve set mathematics back a month.”

Erdos resumed taking amphetamines and did so for every day of his life until his death 17 years later."
https://t.co/rfeiDKtnxa

[XY Han @XYHan_]

Imagine if Erdos were still alive with a GPT-5.5 Pro subscription and amped up on amphetamines.

[Ravid Shwartz Ziv @ziv_ravid]

My two cents on why information theory doesn't quite work in the real world: as someone who's been arguing with people about IT and its connection to ML since 2016(!), I mostly agree with Alex.
(At least) two problems show up when you try to use information theory in ML. First, the distance between theory and reality. Nothing is clean, everything is noisy, full of engineering tricks, and you have to estimate everything. We run experiments with information estimators whose quality we don't really know.
But the bigger problem (and I hope Shannon will forgive me): information theory isn't really about learning. In the classical setting you're in an idealized world where everything is given, with no optimization and no learning. There have been attempts to change this (Stefano Ermon's usable information, for one), and many works took inspiration from IT, but actually applying its tools has had limited success. So, next time that people ask you "Does Information theory 'explain AI you can say, "No, and it doesn't supposed..."

[Dimitris Papailiopoulos @DimitrisPapail]

Information theory is a mathematical theory and a language, not a scientific theory that explains phenomena. Slapping mutual information on all of AI’s mysteries won’t help explain them.

[Ash Jogalekar @curiouswavefn]

The story goes that apparently Born was too afraid of Oppenheimer to show him the petition directly, but had to entertain it to prevent his classes from being boycotted by Goeppert (better known as Goeppert-Mayer later) and others. So one day he invites Oppenheimer into his office and keeps the petition prominently on his table. He steps out for a few minutes making some excuse. When he comes back Oppenheimer's face is white as a ghost's. After that the interruptions (largely) stop.

[Peyman Milanfar @docmilanfar]

don't be disappointed if information theory doesn't explain AI

[XY Han @XYHan_]

In fact, AI replacement fears recently has a lot of parallels with reactions to Wiener's 𝘊𝘺𝘣𝘦𝘳𝘯𝘦𝘵𝘪𝘤𝘴 & Shannon's Info Theory in the 1950s. One optimistic POV is we've been here before. We survived it then; we will survive it now. Again quoting Kline (2015):

[XY Han @XYHan_]

Actually, this is not new. People have tried overextending info theory since the 1950's. History is just repeating itself. Some examples from Kline's Cybernetics Moment (2015):