Terence Tao, a UCLA professor who is considered “the world’s greatest living mathematician,” compared ChapGPT’s o1 inference model to an “average but not completely incompetent” graduate student last month. Graduate students can accurately answer complex analysis problems with ‘many hints.’ And I urge you.”
AI will never be able to beat human teachers, he now tells The Atlantic. “One of the key differences between graduate students (today) and AI is that graduate students learn. If you tell the AI that its approach isn’t working, it may apologize and temporarily correct course, but sometimes it just goes back to what it tried before.”
The good news for math prodigies is that AI and mathematicians will always be more likely to be collaborators, and rather than AI replacing math geniuses, it will enable them to explore large-scale problems that were previously inaccessible, Tao said. I would like to add: “You might have a project and say, ‘Why not try something like this?’” says Tao of the future. And instead of spending hours trying to actually make it work, GPT guides you to do it for you.”
