OpenAI announces ten advances in math and theoretical computer science using AI
OpenAI has unveiled ten significant advancements in mathematics and theoretical computer science, solving problems that had remained open for over a decade. The AI research behind these solutions cost approximately $2,000 in token expenses.
OpenAI, a leading artificial intelligence research company, has announced ten breakthroughs in mathematics and theoretical computer science. These advancements address problems that had been unsolved for at least ten years, with most remaining open for even longer.
The computations for these breakthroughs were performed using an internal version of OpenAI's next-generation model, Astra. The cost associated with generating the solutions, calculated via Sol API's rates, was approximately $2,000 in token expenses.
OpenAI stated that attribution should accurately reflect the research process. The company believes that presenting a proof generated entirely by an AI system as solely human work would diminish the system's contribution and distort the nature of genuine human intellectual labor.
Human researchers assisted in drafting the manuscripts and formalized the proofs in the Lean language to verify their correctness. However, the mathematical arguments themselves were generated by OpenAI's systems. The reported advances include new upper bounds for high-dimensional sphere packing, exponential improvements in binary and spherical codes, and the proof of existence for non-sofic groups.