François CholletVerified account

Knowing the algorithm != Knowing how/why it works. My recommendation: do Kaggle competitions, and use more than neural nets. Use many different ML models. Make it visual. Plot your features.

It's perhaps clearer with a simpler example: implementing SVD teaches you how to implement SVD. Once you're done, you still don't know what it does. Using SVD on various datasets and plotting & using the results is what gives you intuition about SVD. Also, learning math

1/3 I think you raise some good points, but in my view these things are orthogonal. There are many difficult, real world problems where current paradigms like deep learning are inadequate, and cookbook cookiecutter approaches won’t work.

2/3 Knowing what’s under the hood not only makes it easier to debug errors, it also gives me more confidence to modify extend the existing paradigm to tackle new types of problems.