Atomistic Simulations of Oxide-Water Interfaces using Machine Learning Potentials