讨论 Astra 之后的模型能做人类数学家做不到的题
作者引用内部模型说法，认为机器在数学上超越人类不算冒犯，也理解 OpenAI 对公开长期未解难题结果格外谨慎。
"...This internal model past Astra is one that can do things that the best mathematicians in the world cannot." Saying that a post-Astra model can do math that no human mathematician can do feels pretty natural to me. But for a lot of people, that idea is almost offensive. That’s also why I can understand OpenAI being extremely cautious about releasing results on problems that have been open for decades. People are already losing their minds over Navier-Stokes because scientific prestige is deeply tied to human achievement. But practically, I don’t see the point in treating this like a competition. A machine doesn’t sleep, eat, get tired, feel pain, or lose focus. If it eventually becomes better than the best humans at some fields, that’s not an insult to humans. It’s just what the technology was built to do. https://x.com/rohanpaul_ai/status/2100144714251198466/video/1