I see an element of competition in the current situation. On one side are LLM’s with the colossal knowledge of pretty much anything humans ever invented, and the massive compute power behind it to search and combine this knowledge. On the other side are humans, each with limited knowledge and compute power, but possessing something very precious: intuition for beauty and elegance. These have guided human explorers through the vast ocean of potential constructions to use and theorems to prove, allowing for some brilliant discoveries in the past.
The Ultimate Question is whether the colossal knowledge and massive compute power can win over human intuition in all areas.
Here is why I think this is indeed the ultimate question. We are paid to do our research and to educate the next generation of researchers because some of our output is fundamentally useful for society. Continuing our journey we certainly face some big changes in the near future, and thinking of how to adapt our profession is essential. I read some deeply insightful opinions on human-scale mathematics, the importance of human knowledge and understanding, or the way new research makes it into core knowledge of humanity by settling in useful human-edited textbooks. Such thoughts are very important contributions but: they seem to exist under the assumption that the answer to the Ultimate Question is NO, hence humans will still be needed in the future.
If the answer to the Ultimate Question is YES then, I think, all of this insightful guidance becomes irrelevant. If LLMs get better than humans in all aspects of mathematics, then there is no reason for us to be paid anymore. The machine will be able to do it faster, better, cheaper; our profession will degrade into an exotic art that benefits a teeny-tiny proportion of society who find it interesting for its own sake. Funding will be squeezed accordingly. The good(?) news is that we won’t be the only ones. If LLMs can do better than humans with all of mathematics, I expect them to do better in all other areas of science and engineering as well. And the LLM designing engineering structures/clever gadgets/medications/healthcare methodologies/financial or climate models/etc. won’t need our great textbooks with the core of mathematics nicely organised in them. It will have the compute power to go through the ocean of mostly useless theorems spit out by the mathematics LLMs and pick the relevant ones for its actual job.
I’m still optimistic and think that the answer to the Ultimate Question is NO, at least for many years to come. There certainly have been spectacular results by machines but I don’t feel like this applies to all problems in mathematics. My (limited) understanding is that LLM’s are, at the moment, very far from developing a useful sense of beauty and elegance in our fields. And without that, it is really difficult to efficiently navigate the vast ocean of potential theorems waiting to be proved.
Received 24 September 2026.
Add to the discussion