Tackling the Curse of Dimensionality with Physics-Informed Neural Networks