The Mystery of OpenAI's Millennium Problem Breakthrough... Allegations of Research Precedence Emerge
Solved a Century-Old Mathematical Problem,
But Anthropic Was Also Poised to Announce Results
OpenAI Faces Research Precedence and Data Access Allegations
OpenAI announced on September 8 (local time) that it had solved a mathematical problem that humanity had been unable to crack for about 100 years, using its AI chatbot service, ChatGPT. However, as it emerged that a researcher from Anthropic, a rival company, was also about to announce similar research results, suspicions were raised regarding research precedence and data usage.
The "Navier–Stokes problem" solved by OpenAI is one of the "Millennium Prize Problems" selected by the Clay Mathematics Institute in 2000. It concerns determining whether solutions to the equations describing the movement of fluids, such as air and water, fail under certain conditions. A prize of 1 million dollars was offered for solving this problem.
To tackle this challenge, OpenAI employed an internal AI system more powerful than its recently introduced model, GPT-6 Astra. The company stated that approximately 10,000 AI agents were operated simultaneously and found a solution within 88 hours. Thereafter, GPT-6 Astra spent around 17 hours verifying the solution. OpenAI concluded from this proof that, under certain conditions, the equation fails. It explained that a phenomenon known as "blowing up," where the velocity becomes infinite, can occur when calculating the flow of water or air.
If this achievement passes mathematical peer review, it is expected to become a symbolic case demonstrating that AI can not only assist human researchers, but also directly solve the most difficult unsolved problems. OpenAI stated that it would not be claiming the prize money for solving the problem.
According to foreign media including Axios and The Guardian, the issue of research precedence arose immediately after OpenAI's announcement. Tristan Buckmaster, a mathematician at New York University, claimed that upon learning that he and a research team from Anthropic were about to announce a similar breakthrough, OpenAI accelerated its own research work.
In particular, it was revealed that unfinished research works by Buckmaster and Levent Alfoege, a researcher affiliated with Anthropic, were both stored in OpenAI’s coding tool Codex. Both were working on the same problem using OpenAI’s services when OpenAI announced that it had solved the issue. OpenAI has denied allegations that it directly used or accessed the materials. However, it did not rule out the possibility that data generated as the two researchers used OpenAI products may have contributed to improving its own models.
Hot Picks Today
"I'm Scared, Please Don't Go to Japan Right Now"... Tourists on High Alert as Deaths Occur and Trains Halt
- "Thanks to You, I Raised My Kids Well": Farewell Letter to 450,000 km Porter... Hyundai Motor Company Says, "We've Found the Owner!"
- "Receiving 3.8 Billion Won and Claiming 'High Pay Prevents Bribes'... The Country with the World's Top-Paid Prime Minister Plans 64% Salary Hike"
- Seoul National and Yonsei University Early Admission Competition Rates Fall... Korea University Sees Slight Increase
- "300 People Doing This at a Nightclub?"... A Twist Born from Loneliness and Digital Fatigue
This controversy has emerged amid intensifying competition between OpenAI and Anthropic. Since launching ChatGPT in 2022, OpenAI has led the generative AI market, but this year it has been assessed as having ceded technological leadership to Anthropic. It has also been reported that both companies are close to going public. Anthropic's corporate value recently soared to 965 billion dollars (about 1,310 trillion won). It is expected that the company could be valued at up to twice that amount at the time of IPO. OpenAI was valued at 852 billion dollars (about 1,157 trillion won) during its investment round last March.
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.
