We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
OpenAI Blog·Feb 13·5 min read
GPT-5.2 derives a new result in theoretical physics
TL;DR
Theoretical physics lacks efficient methods to compute gluon amplitudes—mathematical expressions describing how gluons (force-carrying particles) interact. GPT-5.2, a large language model, autonomously derived a novel formula for gluon amplitudes that was subsequently formally proved and experimentally verified by OpenAI and academic partners.
✦ Why It Matters
Engineers can explore using large language models for automated scientific hypothesis generation in physics and mathematics, with formal verification as validation.
Key Takeaways
- GPT-5.2 generated a novel gluon amplitude formula without explicit training on physics derivation.
- Formal mathematical proof and experimental verification confirmed the AI-proposed formula's correctness.
- Language models can contribute original discoveries to theoretical physics research.
- This approach reduces manual computational burden in particle physics formula derivation.
Full Summary
Gluon amplitudes are mathematical formulas in quantum chromodynamics (the theory governing strong nuclear forces) that predict how gluons interact and scatter. Deriving these amplitudes manually is computationally intensive and error-prone.
GPT-5.2, OpenAI's large language model, was applied to this problem and generated a previously unknown formula for gluon amplitudes. The model's output was then rigorously validated through formal mathematical proof and experimental verification by OpenAI researchers and academic collaborators.
This represents a concrete instance of AI discovering new physics rather than merely optimizing existing methods. The result demonstrates that language models can contribute to theoretical physics research by proposing novel mathematical relationships that humans subsequently validate.