Abstract: Results from machine learning are notoriously difficult to interpret and verify, making them less attractive for applications in pure mathematics. We will introduce reinforcement learning (RL), a subfield of machine learning, and explain how to apply it to problems in knot theory. This technique produces verifiable results which are sometimesinterpretable by domain experts. We will illustrate this by applying RL to the unknot recognition problem, and to the problem of detecting whether a knot is ribbon. The latter was used to rule out many proposed potential counterexamples to the smooth Poincare Conjecture in 4D. The presentation is based on https://arxiv.org/pdf/2304.09304.pdf.