Technology
Mathematicians confront rapid AI advances amid unverified proof claims and credit disputes
Wired reports that recent machine learning breakthroughs in complex mathematics have sparked apprehension alongside disputes over research credit.
The short version
- OpenAI announced it deployed tens of thousands of automated agents to resolve a 90-year-old math problem associated with a $1 million prize.[Wired]
- Mathematician Tristan Buckmaster claimed OpenAI accelerated its project upon learning of his work with Anthropic researcher Levent Alpöge and sought to sway credit attribution.[Wired]
- Anthropic announced its Claude model proved 29,500 smaller theorems during the formalization of Fermat's Last Theorem, following prior OpenAI announcements of progress on 10 long-standing problems.[Wired]
- The full validity of OpenAI's prize problem solution remains unknown pending formal independent verification by the mathematical community.[Wired]
Key facts
- OpenAI announced it solved a 90-year-old mathematical problem tied to a $1 million prize by deploying tens of thousands of AI agents.[Wired]
- Mathematician Tristan Buckmaster alleged that OpenAI raced to solve the prize problem after becoming aware of his research with Anthropic's Levent Alpöge and attempted to sway attribution.[Wired]
- Anthropic reported that its Claude model proved 29,500 minor theorems to formalize an existing proof of Fermat's Last Theorem.[Wired]
- OpenAI previously announced in August that it achieved advancements on 10 other long-standing mathematical problems.[Wired]
- Researcher Alex Townsend utilized ChatGPT to assist in solving a numerical linear algebra problem that had persisted for decades.[Wired]
What remains uncertain
- The mathematical proof announced by OpenAI has not yet undergone independent verification.[Wired]
Sources
Outlet counts describe coverage, not independent confirmation. Reports may share a wire service or original source.