Codex Helps Rewrite Equations That Stalled Black Hole Plasma Simulations for Decades

Chi-kwan Chan, an astrophysicist on the Event Horizon Telescope team, is using OpenAI's Codex to crack a problem that has limited black hole plasma simulations for decades: the computational cost of tracking every spiral of trillions of particles.

The Spiral Problem: Why Supercomputers Fail at Black Hole Plasma

Plasma around supermassive black holes gets so hot and diffuse that electrons and ions rarely collide. Instead of behaving like a fluid, each particle corkscrews along magnetic field lines. Standard simulations must calculate every minuscule turn, forcing computers to take absurdly small timesteps.

“For decades, this has limited how realistically we can simulate black hole plasma,” Chan said. Even the world's fastest supercomputers spend most of their cycles on these tiny motions rather than the large-scale physics scientists actually care about.

Codex as Algorithmic Lab Assistant: Testable Proposals

Chan suspected that changing the mathematical formulation — rewriting the differential equations so the simulation no longer has to follow every spiral directly — could break the bottleneck. But exploring the space of possible algorithm modifications by hand would take forever.

So he turned to Codex to generate candidate schemes and test them against known solutions. Many failed. “But that’s okay,” Chan said. “Most scientific ideas fail. What matters is that these algorithms are testable.” Unlike black-box AI systems that return answers without showing their work, Codex proposes numerical methods that Chan can inspect, verify, and understand physically.

What Comes Next: Simulating Trillions of Particles

If the approaches Chan is testing with Codex succeed, the new algorithms could eventually let researchers simulate trillions of particles around black holes — enabling study of physics that has remained out of reach since the EHT first imaged M87's black hole in 2019.

Chan sees AI not as an oracle but as a tool for faster iteration, precisely because science demands rigorous, repeated testing. “We don’t accept an idea because it came from Einstein, from a bright student, or from an AI model,” he said. “We accept it only after repeated testing.”

Source: How an astrophysicist uses Codex to help simulate black holes
Domain: openai.com