We're gonna need a lot more mathematicians
- Source
- terrytao.wordpress.com
- Author
- srcreigh
- Date
Cryptographer Amit Sahai wrote a guest post on Terence Tao's blog titled "We're gonna need a lot more mathematicians." His argument is that frontier AI is already producing original mathematical ideas, not just solving posed problems, and at a pace humans cannot keep up with. The bottleneck he identifies is understanding and verification: someone has to check AI-generated results and make them part of the field. His proposed response is to train many more mathematicians for that work rather than fewer. The post drew more than 200 points and nearly 300 comments on Hacker News. It extends an argument this Brief has tracked from other directions, including the rising share of math preprints that acknowledge AI help, and it frames verification capacity as the scarce resource.
Context
Faster AI discovery could increase the need for people who can understand and scrutinize unfamiliar ideas. That is the argument of computer scientist Amit Sahai in this guest essay. He proposes sustained support for research groups that study consequential AI-generated discoveries, including expertise independent of the organization proposing a technology.
His hypothetical fusion-plant example explains the stakes: a mathematical guarantee depends on a model, and people still need to understand that model's assumptions and supporting experiments before making a public decision. This complements the earlier discussion of human mathematicians' role. Sahai's case is for investing in human understanding and agency, rather than a measured forecast of future hiring.
Checking sign-in…
Loading comments…