OpenAI says it cracked 90-year-old maths problem in 88 hours
OpenAI solved a complex Navier-Stokes math problem using 10,000 AI bots in 88 hours, consuming 130 billion output tokens and costing around $10m.
Stock photo for illustration only, not from the actual event
- OpenAI deployed roughly 10,000 AI agents to solve the Navier-Stokes math problem in 88 hours.
- The bots exchanged nearly 3 million messages and used 130 billion output tokens.
- The computational effort equates to an estimated cost of around $10m or £7.3m.
- The solution has not yet been independently verified or accepted by the Clay Mathematics Institute.
OpenAI announced a major milestone in artificial intelligence capabilities, stating that it has found a solution to a decades-old advanced mathematics problem in a matter of hours. By focusing a group of approximately 10,000 autonomous AI agents, the ChatGPT maker tackled a notoriously difficult aspect of the Navier-Stokes equations, which describe fluid motion and turbulence.
The company explained that after beginning training on a new math-adept model at the end of August, it heard rumors on September 1st that two Millennium Prize problems had been resolved. This prompted the firm to deploy thousands of internal AI bots to test solutions on the remaining challenges.
Stock photo for illustration only, not from the actual event
By September 5th, roughly 88 hours after initiating the task, the AI bots successfully reached a solution to the Navier-Stokes existence and smoothness problem. This particular challenge forms the core of turbulence and is tied to a $1m Millennium Prize offered by the US-based Clay Mathematics Institute.
Despite the rapid turnaround time, the scale of computation was immense. The bots exchanged nearly 3 million messages and consumed 130 billion output tokens solely on the Navier-Stokes equations. Based on OpenAI's pricing for its most advanced models, this computational effort would cost approximately $10m, equivalent to £7.3m.
This breakthrough highlights the increasing viability of multi-agent AI frameworks, where thousands of specialized models coordinate and exchange feedback to crack complex scientific puzzles. Nonetheless, independent mathematical validation remains an absolute prerequisite in academia, meaning OpenAI's claims must withstand rigorous peer review before being officially recognized for historical prizes.
The announcement has already sparked discussion within the academic community. Tristan Buckmaster, a mathematics professor at New York University, noted that he and Levent Alpöge, a mathematician at rival firm Anthropic, had also been working on similar solutions. OpenAI maintained that it had no prior knowledge of external work until public disclosures were made.
Source: BBC Business
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment