GPT-6 Astra solves a neuroscience problem left open since 1988 in 100 minutes
A Flatiron Institute neuroscientist gave GPT-6 Astra a long-standing question about how chaotic brain activity is, and after 100 minutes its answer matched simulations, with the paper crediting GPT-6 and Claude.
this might be one of the clearest signs yet that AI can do real science. ai just solved a problem scientists have been stuck on since 1988. back then some physicists built a simple model of brain activity and asked a basic question. "just how chaotic is it?" they couldn't fully answer it. neither could anyone else for almost 40 years. now a neuroscientist at the Flatiron Institute, who honestly thought it couldn't be solved gave it to GPT 6 Astra. it worked for 100 minutes. then its answer matched real computer simulations. even he couldn't really follow how it got there. so he kept working with the ai until he could explain it in a way other scientists could understand and when he published the paper, he openly gave credit to GPT-6 and Claude for helping. ai might finally help us understand our brain.
This post quotes another post: x.com/d_g_clark/status/2108574780957905073