Chi-kwan Chan

We ran a Hackathon on Agentic Coding for Research with a fun group: ~ 25 participants from astronomy, planetary science, physics, and applied math. In 4 hours, we went from an intro to using tools like OpenAI Codex on real problems. By the end, participants shared 10 projects.

Eleven telescopes. Countless scientists, engineers, and operators. One goal: a movie of the black hole in M87. The M87 MOVIE campaign is now underway. Follow along as the campaign unfolds. #M87Movie

Self-play reinforcement learning ...

End of a busy semester: celebrating undergrad Hina Suzuki’s Outstanding Senior Award, then traveling to Phoenix for grad student Ningyuan Xu’s excellent talk at the Texas Symposium. Very proud of both of them.

This is the time of year when we bring our students to the University of Arizona’s High Performance Computing Center. The students got in the server room for an up-close look at the supercomputers that power so much of our research. Always a highlight of the semester!

Wow! Cleo Abram and Mark Rober visited Steward Observatory's Mt. Lemmon/Catalina Station and discussed the Catalina Sky Survey: youtu.be/LOn-mmezykQ

Went to Los Alamos again and talked about AI, multi-physics simulations, and fusion, and stopped by the VLA on our way back to Tucson.

With Nobel Laureate Brian Schmidt ;-)

Returned from OpenAI. Lots of interesting ideas shared on the future of AI in education and scientific discovery.

After our local EHT meeting last week, we stopped by the Arizona Radio Observatory labs. Fascinating to see the cutting-edge instruments up close!

First Images from the NSF-DOE Vera C. Rubin Observatory: youtube.com/live/dF1g-Ru8m… P.S. The 8.4-meter mirror was fabricated at the Richard F. Caris Mirror Lab at UA rubinobservatory.org/explore/how-ru… ;-)

Did you know that black holes are spinning?🌀🔎 To find out just how fast, EHT researchers used AI to connect simulated data to actual EHT measurements of M87 and SgrA*. They found that the black hole in SgrA* spins almost as fast as physically possible! aanda.org/articles/aa/fu…

#A #NeuralNetwork trained on millions of synthetic data sets indicates the Milky Way’s central black hole spins near its maximum rate, with its axis oriented toward Earth. @AandA_journal doi.org/g9pbh9 phys.org/news/2025-06-n…

In python, for i in range(-3,2): print(f'hash({i:2}) == {hash(i):2}') hash(-3) == -3 hash(-2) == -2 hash(-1) == -2 🤯 hash( 0) == 0 hash( 1) == 1 And this is not a bug: #L136" target="_blank" rel="nofollow noopener">github.com/python/cpython…

The very timely Veritasium video was filmed at UA campus: youtu.be/zB_OApdxcno

It was great to return to the department where I earned my degree and present! I enjoyed sharing my thoughts on the connections between physics and AI with the community that helped shape my academic path. BTW, It takes only a few lines of python to implement a Hopfield network.

But I also feel there's a paradigm shift—there are just so many of these tricks now, maybe they're not as valuable anymore… Regardless, one thing is certain—AI is changing physics forever.

It’s clear that there are countless of these theorems or methods hidden in AI models. With my old-school physics training, I think pulling out these reusable tricks, understanding them, and applying them in new ways is very important. ...

AI is now officially physics. 😁 Joking aside, it’s amazing how neural networks can find patterns in data and make them useful. In physics, we discover little tricks that work in multiple situations, then give them fancy names like theorems or methods. ...