AI/ML for Nuclear Theory: Opportunities and Challenges

Beyond their many impacts in our societies, artificial intelligence and machine learning are also often seen as representing a paradigm shift in scientific research. Their unsurpassed ability to learn complex patterns in large datasets have found multiple applications ranging from designing powerful emulators of complex theories, to processing massive amounts of data to opening new ways to solve inverse problems. Nuclear theory offers a unique playground for the application and development of AI/ML techniques. In contrast to other natural science domains or human sciences, nuclear theory has very robust foundations that can be tracked back to the Standard Model of Particles. In fact, while not matching the accuracy and precision of particle physics, the predictive power of nuclear theories is quite good overall. At the same time, all nuclear theories rely on a small set of free parameters that must be calibrated on experimental data, making the theory phenomenological by construction. Finally, most if not all nuclear theories are computationally expensive. In this presentation, I will use selected example of recent applications of AI/ML in nuclear theory to discuss future opportunities and challenges in further extending the scope of these methods to improve our understanding of nuclear systems.