Imagine mathematics as a vast, untapped wilderness of open conjectures and unsolved puzzles. For centuries, these problems have served as the 'fuel' for human discovery, pushing mathematicians to invent entirely new tools and techniques just to crack a single nut. But according to Fields Medalist Terence Tao, we've entered a new era: the age of automated strip-mining.
From Case Studies to Population Studies
Traditionally, math has been a game of intensive 'case studies'—one brilliant mind obsessing over one problem for a decade. AI has flipped the script. By treating math as a 'population study,' AI can sweep across thousands of problems simultaneously. We're already seeing the results; reports indicate that AI tools have helped move roughly 100 Erdős problems into the 'solved' column since late 2025.
The Risk of a 'Net Negative' Solution
Here is the provocative part: Tao warns that open problems are a non-renewable resource. When a human solves a hard problem, they usually leave behind a trail of new theories and insights that benefit the entire field. When an AI solves a problem—especially through formal proof search—it might provide a machine-checkable 'yes,' but without the conceptual bridge that humans can actually use to learn.
If we indiscriminately mine these problems for solutions without extracting the underlying mathematical intuition, we risk destroying the very ecosystem that trains future mathematicians. Tao suggests we might even need to declare certain classes of problems 'off-limits' to AI to preserve their educational and social value.
A New Mathematical Frontier
We aren't just talking about efficiency; we're talking about the survival of mathematical intuition. If the 'low-hanging fruit' of open conjectures is vacuumed up by bots, where will the next generation of human geniuses find the friction they need to grow?
Sources
Media
