I think the key is to stop trying to make AI-generated math compete directly with human mathematics.
AI can generate and formally verify huge numbers of results, but the real status should come from what happens afterward: independent verification, simplification, human explanation, connections to existing work, and eventually actual use by other mathematicians.
So instead of saying “academia should value understanding more,” we need a system where those contributions are actually visible and rewarded.
AI makes producing mathematical truth cheaper. The scarce resource becomes attention and understanding. The challenge is building incentives around that scarce resource.
I think the key is to stop trying to make AI-generated math compete directly with human mathematics.
AI can generate and formally verify huge numbers of results, but the real status should come from what happens afterward: independent verification, simplification, human explanation, connections to existing work, and eventually actual use by other mathematicians.
So instead of saying “academia should value understanding more,” we need a system where those contributions are actually visible and rewarded.
AI makes producing mathematical truth cheaper. The scarce resource becomes attention and understanding. The challenge is building incentives around that scarce resource.