For the past few years, we have been intensively collaborating withmathematicians from Oxford and the University of Sydney. As a product ofthis collaboration, we have successfully demonstrated that analysing andinterpreting the outputs of (graph) neural networks -- trained on carefullychosen problems of interest -- can directly assist mathematicians withproving difficult theorems and conjecturing new approaches to long-standingopen problems. Specifically, using our method we have independently derivednovel top-tier mathematical results in areas as diverse as representationtheory and knot theory. By doing so, we present AI as the mathematician's "pocket calculator ofthe 21st century". The significance of our result has been recognised bythe journal Nature, where our work featured on the cover page.In this talk, I aim to tell you all about our findings, from a personalperspective. Expect key details of our modelling work + an account of howit felt to interact with top mathematicians.