An explanatory article on how can MLP is still useful comparing to modern machine learning architectures without convolutions, nor attention.