Neural Network Training Complexity

John discusses the computational intractability of training neural networks when dealing with certain input distributions, shedding light on the challenges faced despite advancements in deep learning practices. The conversation delves into the complexities of training neural networks under specific input conditions, questioning the feasibility of achieving efficient training methods in certain scenarios.