> his type of bottom up approach is terrible idea for a fast moving area like ML. Of course not. When you learn this way you will realize there are a myriad of

> his type of bottom up approach is terrible idea for a fast moving area like ML. Of course not. When you learn this way you will realize there are a myriad of problems that can be solved with more simple algorithms. Trying to make every problem fit to neural networks is pure cargo culting.