MICDE State of AI: End-to-End Computational Scientific Discovery

This special event will present a recent breakthrough from the groups of Profs. Kai Sun (Physics) and Venkat Viswanathan (Aerospace). Fractional Chern insulators are exotic topological states of matter that host anyons, a proposed foundation for fault-tolerant quantum computing, but their small energy gaps confine them to extremely low temperatures. The team asked frontier AI models to propose a new theoretical framework for topological flat bands, then set a swarm of agents to explore it, coordinating only through a shared message board, with every candidate checked by GPU-accelerated computations. The swarm found a surprise: departing from the “ideal” quantum geometry long considered optimal raised the many-body gap more than tenfold over previous benchmarks and beyond what agents working alone reached. AI-assisted analysis then distilled this into a simple sheared Landau level model that explains the mechanism (reshaped electron orbitals stiffen the short-range repulsion protecting the state) and yields a design index computable without many-body simulation.

Proposing the hypothesis, searching it,  explaining what it found, and presenting a model offers an early look at end-to-end scientific discovery with AI.

Professor of Aerospace Engineering, University of Michigan