Is 'feature extraction' a core machine learning task?

Question

I have been arguing with a friend about 'feature extraction'. He says the main task of ML is to extract features. But I disagree. In a common-sense feature extraction is not an ML task. If we consider wx+b as the simplest way to represent ML, the task of ML is to find the best w and b. x is the feature. ML tries to find out the best w and b values for a given x, it matches with the training data and thus learns how to find w and b.

My friend says it is the core task of ML to extract features. But as I know feature extraction is a data preprocessing task mainly.

Answer 1

Extracting features is an important job in ML. Without features you cannot find the best "w" and "b". If you are able to find w and b without feature extraction then you don't really need to go forward with ML.