I have the utmost respect for Terrence Tao and the mathematics community. I don’t really like the fact that the frontier model vendors are pushing the envelope without first building consensus in the community.
However, I feel the mathematical community is a bit righteous by claiming that model vendors are totally irresponsible in their behavior. These vendors may have legitimately improved the state of the art in mathematics, so what are they supposed to do with the results? Sit on them? Become king and queen makers by choosing the people that should get credit?
The world is irrevocably changed with AI, and I’m glad that the AI vendors are pushing the agenda forward. Yes, they’ve perhaps destroyed the career trajectories of mathematicians who were seeking those results. That said, that is the risk of doing mathematics— there have been innumerable times mathematicians have been scooped.
Yes, the AI proofs are messy. These proofs often do not follow mathematical norms. They sometimes seem like they’re from another “planet”. Like other crackpots (or misunderstood crackpots) before AI, mathematicians can choose to ignore these proofs. Or they can investigate them to see what lies within.
Let the frontier models release what they know; transparency is what we should all strive for. What is a credible proof, who gets credit, and how that credit gets assigned is a social norm that only mathematicians get to decide.
However, I feel the mathematical community is a bit righteous by claiming that model vendors are totally irresponsible in their behavior. These vendors may have legitimately improved the state of the art in mathematics, so what are they supposed to do with the results? Sit on them? Become king and queen makers by choosing the people that should get credit?
The world is irrevocably changed with AI, and I’m glad that the AI vendors are pushing the agenda forward. Yes, they’ve perhaps destroyed the career trajectories of mathematicians who were seeking those results. That said, that is the risk of doing mathematics— there have been innumerable times mathematicians have been scooped.
Yes, the AI proofs are messy. These proofs often do not follow mathematical norms. They sometimes seem like they’re from another “planet”. Like other crackpots (or misunderstood crackpots) before AI, mathematicians can choose to ignore these proofs. Or they can investigate them to see what lies within.
Let the frontier models release what they know; transparency is what we should all strive for. What is a credible proof, who gets credit, and how that credit gets assigned is a social norm that only mathematicians get to decide.