GPT-6 Astra 用 100 分钟解出 1988 年悬而未决的神经科学问题
Flatiron 研究所神经科学家把 1988 年遗留的脑活动混沌度问题交给 GPT-6 Astra，100 分钟后答案与模拟吻合，论文致谢 GPT-6 和 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.
这条是引用着另一条写的：x.com/d_g_clark/status/2108574780957905073