Chi-kwan Chan, a researcher at the University of Arizona and Steward Observatory, is using Codex to help develop algorithms for simulating the plasma that surrounds black holes. The effort aims to reduce computational costs that currently limit how realistically researchers can model particle behavior near the event horizon.
Modeling a spiraling plasma
Black hole gravity is extreme, and light cannot escape once it crosses the event horizon. Scientists study the region just outside that boundary because matter there emits observable light. Chan is a member of the Event Horizon Telescope (EHT) collaboration, which produced the first image of a black hole in 2019 and is gathering observations to produce the first video of a supermassive black hole, focusing on M87.
Simulating the plasma around black holes is computationally demanding. In hotter, more diffuse regions electrons and ions rarely collide and instead spiral tightly around magnetic field lines. Standard particle simulations must resolve each tiny turn, forcing very small timesteps and leaving supercomputers spending most of their time on minute particle motions rather than larger-scale dynamics.
Codex as an algorithmic assistant
Chan suspected that alternate mathematical formulations could avoid following every microscopic corkscrew directly. He turned to Codex to generate candidate numerical schemes and implement them for testing. Codex proposed many approaches; not all were correct, but Chan said the key is that the algorithms are testable. “Most scientific ideas fail. What matters is that these algorithms are testable,” he said.
The group uses Codex to produce implementable schemes that can be inspected and validated. Chan emphasized that scientific acceptance requires repeated testing, regardless of whether an idea originates from a person or an AI model. He argued that science is suited to AI assistance because proposed methods can be verified rigorously.
If the approaches being tested succeed, the new algorithms could eventually allow simulations to follow trillions of particles around black holes, enabling studies of physics that have been out of reach for decades.
Original source: OpenAI News