Math Professor Accuses OpenAI of Copying His Work, Says AI Compute Power Makes Racing to Publish 'Pointless'

Last week, OpenAI announced its AI model had solved a famous math problem after just 88 hours. However, NYU professor Tristan Buckmaster claims OpenAI may have copied his work when they approached him about collaborating. Buckmaster and his colleague, Levent Alpöge, were working on the Euler problem when they independently discovered a solution using AI tools from Anthropic and ChatGPT.

OpenAI reached out to Buckmaster, who questioned the timing of their discovery, noting they had learned of his work only in the past few days. When he asked if their AI systems had been trained on his research, OpenAI declined to answer, leading Buckmaster to doubt their methods. OpenAI stated they do not use user data to solve problems but could not rule out anonymized data use.

Buckmaster criticized OpenAI's decision, asking how it could be ethical to use customers' data to outpace them. Alpöge's affiliation with Anthropic reportedly upset OpenAI, who sought to remove him from authorship. Buckmaster claims OpenAI's Sebastien twice expressed a desire for Alpöge's removal and called the company's actions appalling.

He and Alpöge have more research to publish, but Buckmaster questions the point of racing ahead when AI can perform work that previously took years. He suggests a discussion is needed on how mathematics should operate in this new environment, with academics from OpenAI, Anthropic, and DeepMind discussing ground rules for collaboration.