Meta AI aids work on five open maths questions
Meta says its Muse Spark artificial intelligence models helped mathematicians produce answers to five previously open research questions using the standard Meta AI chat interface, highlighting a new form of collaboration between general-purpose AI and expert researchers. The company published six mathematics papers on October 2 covering probability, differential equations, group theory,…
Meta AI has collaborated with mathematicians to provide answers to five previously unexplored research questions, as seen in six new mathematics papers published on October 2. These papers cover probability, differential equations, group theory, optimization, arithmetic physics, and non-associative algebra. Using Meta's Muse Spark 1.1 and 1.2 models in Thinking Mode, researchers guided the AI, explored ideas, and developed arguments, while separate mathematicians reviewed the work.
The papers identify passages drafted primarily by researchers versus those primarily by AI to maintain traceability. One paper, by Aykut Arslan, explores the threshold for fitting random Gaussian points in high dimensions to a centered ellipsoid, identifying a sharp transition at which the probability of such an ellipsoid existing shifts.
Another paper, led by Leonard Dinh, proves finite-time collapse in a specific type of differential equation. Joseph Phillip Brennan and Milana Golich used Muse Spark to disprove a conjecture about finite semiabelian groups, finding a counterexample with 384 elements. Meta acknowledges that the separate AI agent Nilradical reported a different counterexample on September 16, highlighting the need to carefully assess the broader claim of solving open problems.
Written by urgent.news from Arabian Post's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
