Artificial Intelligence for Solving General Conformal Field Theories

The use of artificial intelligence (AI) in physics is burgeoning. For a few decades, machine learning has been used to tackle classification and regression problems in particle physics. More recently, AI has made an impact on the study of conformal field theory (CFT) in both applied and theoretical contexts. In particular, very recently, neural networks utilizing reinforcement learning have been applied with success to the conformal boostrap program to solve for the fundamental data of 2D CFTs.

We will build on such techniques and apply them to a wider range of problems. We will examine how neural networks using soft actor-critic reinforcement learning algorithms can be employed in the context of existing numerical conformal bootstrap methods. In addition, we consider how the techniques of AI can be applied more generally, providing potentially significant advantages in solving higher dimensional or non-unitary CFTs that cannot be handled by traditional methods in conformal bootstrap. Finally, we will discuss future avenues for the application of AI to CFTs.