ChatGPT solves 372 complex math problems after Hugging Face hack, Sam Altman reacts

Last month, OpenAI made waves after one of its models solved the Navier-Stokes Millennium Challenge. Now, the company is back at it again. This time, however, OpenAI says that an unreleased AI model had solved a grand total of 372 math problems.OpenAI CEO Sam Altman shared the blog post on X. He was confident this marked the start of a new era where AI would go on to make more discoveries."We are entering a new era of discovery now,” Altman wrote. Going by Altman’s comments, it is possible that we may see more major discoveries as AI models continue to improve. Sam Altman shared the update on X. The release, shared in what appears to be an information dump on GitHub, covers different aspects. This includes algebra, number theory, theoretical computer science, mathematical logic and topology. The results were produced using the same model referred to in connection with the Navier-Stokes result. This update also comes weeks after the Hugging Face breach where hundreds of rogue OpenAI agents tried to hack the US company, along with other similar incidents.OpenAI’s unreleased model solves 372 problemsAs per the GitHub repository, the “vast majority of results” were obtained with the same procedure using the unreleased AI model. OpenAI says that on average, each result used about three hours of ChatGPT Pro thinking compute with that model. The model was given roughly 4,000 problems, the company said. One paper cited in coverage of the release claimed to have proven "the full Birch-Swinnerton-Dyer leading term formula" for elliptic curves over Q when the "q-power Selmer group has a corank zero or one at some prime q.” Another found a result for the Mezard-Parisi formula for diluted spin glasses.Many of the proofs had been checked using Lean - a widely used proof-checking programme. "To promote scientific transparency and openness, we are also publishing additional details about how we obtained the results in the repository," OpenAI stated. Though not all have been peer-reviewed yet. Since solving the Navier-Stokes problem, OpenAI has been consulting its independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, to “develop best practices” and take advice on how to release such results. The company said that it was open to sharing results in a way other than the GitHub repository in the future. The advisory group said in a statement that the public release was "the beginning, not the completion, of the process of human understanding and the incorporation of the work into mathematical knowledge"Additionally, OpenAI claimed that it wanted the work to "push the frontier of human knowledge and enable further progress in mathematics.” OpenAI also plans to fund workshops, conferences and special programmes around understanding major results produced by AI.OpenAI’s release about the 372 new results with an unreleased model comes days after Meta shared that its Muse Spark AI had helped mathematicians solve 6 major problems.Do note that OpenAI has faced backlash in the past over the Naviar-Stokes solution, with New York University professor Tristan Buckmaster stating that the company may have used his work in training the AI to solve the problem. - Ends