Awesome article, and awesome analogy, except using machine learning algorithms is not as “abstract” and as simple as using a microwave :) We need lot of works to be done in Machine learning frameworks to be as abstract as possible. Currently every machine learning algorithm has tons of hyper-parameters and performs poorly if you don’t tweak these parameters. It’s like having a microwave with 1000 knobs that won’t cook your food unless you dial the correct settings on all the knobs :)

An easy enough to use abstraction would be a framework with just 2 methods: learn(inputs, outputs) and predict(inputs). But currently, we’re very far from this…

Cheers!