The first AI simulation of a black hole

TL;DR

This paper presents an AI-based simulation approach for black hole physics, demonstrating that deep learning models can predict accretion flow dynamics significantly faster than traditional methods while maintaining reasonable accuracy.

Contribution

It introduces a novel application of deep neural networks to simulate black hole accretion flows, outperforming traditional numerical solvers in speed.

Findings

Deep neural networks effectively learn black hole accretion physics.

AI models evolve accretion flow faster than traditional methods.

The approach maintains reasonable accuracy over extended periods.

Abstract

We report the results from our ongoing pilot investigation of the use of deep learning techniques for forecasting the state of turbulent flows onto black holes. Deep neural networks seem to learn well black hole accretion physics and evolve the accretion flow orders of magnitude faster than traditional numerical solvers, while maintaining a reasonable accuracy for a long time.