Tackling the curse of dimensionality with physics-informed neural networks