OpenAI Unveils Major Breakthroughs in Mathematics, Physics, and CS
AI models have solved mathematical hypotheses and challenged fundamental computer science theories.
New Hypotheses and Their Impact on Science
OpenAI has once again surprised the scientific community. Just weeks after publishing methods to solve 372 complex mathematical problems, the AI laboratory has presented new, large-scale discoveries. This time, the results achieved by artificial intelligence span mathematics, computer science, and physics.
Two new documents are now available on OpenAI’s GitHub page. One addresses the Sogge local smoothing conjecture in wave equations, which was previously considered an unsolved mathematical puzzle. Even more significant is the second discovery: the model proved that the Exponential Time Hypothesis (ETH) is false. ETH is considered a fundamental assumption in computer science, suggesting that certain tasks cannot be solved quickly. If ETH is indeed false, it will enable scientists to create significantly more efficient algorithms in fields such as databases, biology, and geometry.
Collaboration with Experts
OpenAI has adopted a different approach this time. To analyse the findings in physics and computer science, the company invited groups of independent experts. While the mathematical results are publicly available for any researcher to verify, experts were tasked with creating their own explanatory materials for the computer science results, and the physics findings were handed over entirely to relevant scientific groups.
Notably, the very scientists who proposed the ETH hypothesis years ago have been tasked with analysing the results generated by OpenAI’s models. In the field of physics, the work involves proving quantum field theory in four dimensions, which remains one of the most difficult challenges for physicists.
Academic Integrity and the Future
OpenAI states it does not intend to claim academic credit for these discoveries. All scientific publications and materials will list the experts and researchers who participated in interpreting the results as authors. The company says this process aims to strengthen the role of artificial intelligence in science and increase researcher engagement.
For scientific circles in Georgia, this news is significant as it demonstrates that artificial intelligence is no longer just a tool for writing text or generating images. It is becoming a full-fledged partner in fundamental science, capable of helping humanity solve problems that have remained unresolved for decades.