“NVIDIA's key insight regarding AlexNet was that it represented a deep learning approach capable of learning any function, making it a universal function approximator.”