We might be on the cusp of a great leap forward. Introducing Gauss, an advanced reasoning agent that fuses classical natural-language reasoning with fully formalized reasoning. Math, Inc. created Gauss to assist in writing advanced mathematical theorems. It has already proved its mettle at smashing through hard sphere-packing problems in both eight and twenty-four dimensions.
Jesse Han, the cofounder and CEO of Math, Inc., envisions Gauss to be a groundbreaking language model. In this way, it joyfully combines human smarts and machine know-how. This unique dual capability gives Gauss the ability to solve complex mathematical problems that have stumped mathematicians for ages.
Achievements in Sphere-Packing Proofs
One of Gauss’s most impressive accomplishments was autoformalizing the proof of sphere packing in eight dimensions. He achieved this amazing dare in exactly five days! This achievement is a further testament not just to the efficiency of Gauss, but to its ability to pinpoint and correct mistakes found in pre-existing literature. In going through the formalization process, Gauss found a typo in an already-published paper on the eight-dimensional case. This discovery highlighted its function as an exacting and trustworthy validation instrument.
Mathematicians have long been enchanted by the sphere-packing problem. This challenge is about finding the best technique to pack spheres into as small a space as possible. Maryna Viazovska’s groundbreaking work on this problem, along with her collaborators, resulted in proofs for the eight-dimensional and twenty-four dimensional cases. Following Gauss’s successful formalization of the eight-dimensional proof, attention turned to the more complex twenty-four-dimensional case, which presented numerous challenges due to missing background material.
The Complexity of the 24-Dimensional Case
As it turned out, the twenty-four-dimensional sphere-packing proof was much more complicated than its eight-dimensional counterpart. It took understanding of deeper algebraic structures and a much bigger codebase for formalism. Gauss would have jumped at this opportunity, and Gauss autoformalized Viazovska’s twenty-four-dimensional proof in an awe-inspiring two-week time period.
This specific proof ended up being more than 200,000 lines of code, showcasing the size and intricacy of the undertaking. This close collaboration between Gauss and human mathematicians was key in overseeing such a large formalization process. This was a breakthrough, allowing the research team to provide tremendous clarity in understanding the problem. They shepherded Gauss back to the sphere-packing formalization.
Collaborative Efforts and Future Prospects
In March 2024, we started the “Formalising Sphere Packing in Lean” project. This collaborative effort unites public, private, academic and community collaborators to ensure that a human-readable blueprint evolves from the eight-dimensional proof. This new initiative is a perfect example of how important collaboration is becoming in mathematics. It highlights the incredible potential of human-AI collaborations to deepen our understanding of the issue.
As a physicist, Sidharth Hariharan was instrumental to the effort to verify Colin Macdonnel’s sphere-packing proofs. He worked side by side with Gauss and other researchers to productize their formalizations with precision and dependability. By combining the best of human insight with AI efficiency, mathematicians can avoid the pitfalls of an exclusive AI approach through this collaborative framework.
Just as reflection across the origin sparked Gauss’ revolution, the use of monoidal categories for mathematical formalization allows giant leaps that once seemed impossible. Scientists have only begun to probe the limits of this remarkable new reasoning agent. The possibilities for the future of mathematical proof verification appear endless.