AI Breakthrough: ChatGPT Acts as Co-Author for Significant Theoretical Physics Discovery

Theoretical physicists have achieved a historic milestone by using artificial intelligence to solve a mathematical problem that had eluded human collaboration for decades. Harvard professor Andrew Strominger and his team leveraged OpenAI's specialized tools to prove that certain gluon amplitudes are nonzero. This discovery marks one of the first times an AI has served as a creative co-author in a significant theoretical physics breakthrough. The research focuses on the complex probabilities of particle interactions involving gluons, which are the particles responsible for binding atomic nuclei. While standard physics intuition suggested these specific amplitudes should vanish, the team suspected otherwise but could not find the mathematical proof. The project successfully utilized ChatGPT-5.2 Pro and an internal model called Super Chat to derive the solution.

A Fifth Collaborator

Super Chat acted as a powerful fifth collaborator by reasoning through the problem for twelve continuous hours. It eventually proposed a piecewise-constant closed-form expression that confirmed the physicists' suspicions. The human team spent the following week manually verifying the calculations before publishing their findings as a preprint on arXiv. Strominger, who was initially skeptical of early AI models, described the experience as exhilarating and profoundly different from using a simple machine. He noted that there was a specific moment where the AI felt like a creative person rather than a calculator. This shift suggests that the technology is moving beyond data processing toward genuine symbolic reasoning.

Retooling for the Future

The success of this collaboration highlights the growing role of specialized AI in high-level academic research. Programs like OpenAI for Science are hiring experts to train models specifically for math and physics. Scientists like Alexandru Lupsasca believe this represents the most significant change in theoretical physics in decades. Despite the impressive performance of the AI, Strominger maintains that human scientists remain irreplaceable. He argues that researchers must constantly retool their skill sets to integrate new technology. This partnership demonstrates how AI can empower physicists to solve intractable problems rather than simply replacing human thought.