Shafaqna English- OpenAI has announced that an advanced experimental AI model has generated solutions to 10 longstanding open problems in mathematics and theoretical computer science, marking a significant step in the use of artificial intelligence for scientific research.
The company said the breakthroughs were produced by an internal version of its next-generation model, Astra, during research evaluations. The AI generated mathematical arguments that were later reviewed, organized into research manuscripts by human collaborators, and formally verified using the Lean proof assistant.
The reported advances span a wide range of fields, including high-dimensional geometry, coding theory, group theory, arithmetic circuit complexity, quantum computing, lattice cryptography and graph theory. Among the claimed results are new bounds for sphere packing and coding problems, the construction of non-sofic groups, a disproof of Connes’s rigidity conjecture, advances in quantum parallel repetition, and solutions to several longstanding problems posed by mathematician Paul Erdős.
OpenAI said the computational cost of generating the solutions would be roughly equivalent to $2,000 using its current API pricing. The company also released AI-generated explanations describing the reasoning process behind each result and formal proof certificates for public review.
The announcement follows OpenAI’s recent launch of ChatGPT for Academic Researchers, a program that aims to provide free access to advanced AI models for 100,000 researchers by 2027. The company said it believes AI should serve as a collaborative research tool while emphasizing that attribution should accurately reflect the role AI systems play in generating scientific discoveries.
Source: OpenAI