Terence Tao

I love to read about Terence Tao's incredibly insightful experience reports with gen AI, in particular with [chat](chat)GPT and now the new o1-preview model (https://lnkd.in/gAjs2N7K). His goal when using an AI is to support him in solving sometimes ill-posed, advanced mathematics problems. He reinforces the point that, while genAI is improving significantly (and may, within a couple of iterations, truly support mathematical discovery), it does not know and cannot know at this point what really matters:

"Here the results were better than previous models, but still slightly disappointing: the new model could work its way to a correct (and well-written) solution if provided a lot of hints and prodding, but did not generate the key conceptual ideas on its own, and did make some non-trivial mistakes."

The "key conceptual ideas" come from the human. I have tried to make this point before: among the gazillion things the AI could try to prove (and will generally be able to prove in the near future), only a few really matter, have deeper meaning and take you to the solution. Scientific discovery is at some level a walk in a terribly infinite combinatorial space, an infinite garden of forking paths, and it turns out that the really important forks require human sensemaking. For the rest of the walk, the AI will do.

I can't resist illustrating the humanness of mathematics with this picture of 10-yr-old Terence Tao with Paul Erdös (1985).