Artificial intelligence (AI) is revolutionizing both the technological industries and fundamental sciences. While breakthroughs in long-standing and more applied problems such as playing Go or protein folding have been well-publicized, important progress has been achieved in theoretical physics through AI techniques. For example, it provides a new way for simulating lattice QCD with matter in four dimensions, or to address issues related to building string compactifications and string field theory, while also solving difficult mathematical questions (such as building a Calabi-Yau metric or solving the accessory parameter problem in Poincaré uniformization for Riemann surfaces). On the other hand, AI algorithms are notoriously resource-consuming and their results are difficult to interpret, which is a major hurdle for scientific discovery. Physics provides a natural framework to construct an effective theory of learning and improve our understanding of AI and its performances. The objective of this interdisciplinary workshop is to bring researchers of these different areas together to improve the transfer of knowledge, techniques and ideas in order to provide deeper insights into fundamental aspects of our Universe, the scientific discovery process, and artificial intelligence.