A New Kind of Research Partner
Fields Medalist Terence Tao shared a public ChatGPT conversation, later discussed on Hacker News, showing how he used the model while exploring a potential counterexample related to the Jacobian Conjecture. In the log, Tao directs the inquiry — feeding the model algebraic expressions and high-level concepts — while the AI performs symbolic manipulations and calculations.
Tao described the AI's role as a "highly proficient, tireless, and fast" assistant, handling tedious but error-prone computational steps rather than generating the mathematical ideas itself.
The AI's Role: Assistant, Not Author
Tao provided the intellectual framework and strategic direction; the model executed specific, well-defined tasks. According to the shared log, these included:
- Performing symbolic algebraic expansions.
- Verifying multi-step calculations.
- Flagging potential inconsistencies in the reasoning.
- Translating concepts into LaTeX notation.
This mirrors a pattern already common in AI-assisted coding — the human directs strategy, the model handles execution — now extended into abstract mathematics, a domain without the rigid syntax of programming languages.
Why It Matters
The exchange is a concrete example of a top mathematician using an LLM as a computational aide rather than an autonomous solver. It does not represent progress toward solving the Jacobian Conjecture itself, but it illustrates one way current models are being incorporated into expert research workflows.