GPT-6 Sparks New Scientific Revolution: AI Independently Discovers New Mathematical Theorem for the First Time

OpenAI this week released GPT-6, and the model's performance in scientific exploration has drawn global attention. In mathematics, GPT-6 independently proved a new theorem regarding prime number distribution — verified by three human mathematicians, becoming the first mathematical theorem discovered and accepted by AI in history.

Capability Leap

GPT-6 demonstrated breakthrough capabilities in three key scientific domains:

Mathematical Theorem Discovery: GPT-6 can not only verify proofs but independently formulate new mathematical conjectures and provide proof pathways. An MIT mathematics professor commented: "This is the most important moment in mathematics since computer-assisted proofs."

Protein Structure Prediction: The combination of AlphaFold 3 and GPT-6 has pushed protein design to new heights. GPT-6 can inversely design protein structures based on functional requirements, successfully aiding the design of three potential anti-cancer proteins.

Materials Science: GPT-6 helped discover two novel superconducting materials, with related papers already published in Nature.

Safety Boundary Discussions

Significantly, GPT-6's capabilities have touched the boundaries of existing AI safety frameworks. The Center for AI Safety has called for governments to urgently reach consensus on "open boundaries for AI scientific research."

Industry Reaction

Google DeepMind announced Gemini 2.0 is under urgent development, while Anthropic stated Claude 4 will focus on AI interpretability. The open-source community expressed strong dissatisfaction with GPT-6's closed-source strategy, with Meta announcing accelerated Llama 4 development.