RE: LeoThread 2026-05-26 17-43

There appears to be a growing recognition that humans are poor at mathematics.

A gibbon would ridicule an Olympic climber—the human body isn't optimized for climbing. Similarly, accumulating evidence suggests the human brain may be far from ideal for advanced mathematics.

This isn't meant as a slight toward mathematicians; even highly accomplished people can acknowledge that others are far more capable. Mathematics seems to sit midway on Moravec’s paradox—between chess (where machines surpassed humans some time ago) and cooking (which likely remains years away for general AI). Mathematics is relatively difficult for humans, so machines seem poised to overtake human ability in this area.

AI mathematics still has significant weaknesses. For example, AI systems have not yet demonstrated the capacity to identify promising research directions or originate new foundational concepts. Still, they are becoming superhuman in certain respects, and when AI reaches superhuman levels in a domain, rapid and far-reaching change typically follows.