Just a thought copied from somewhere

Why should we learn #ml from scratch rather than just using libraries??

Let’s take an example:

Manager: We need a system for detecting pedestrians in real-time on the road!

Eng: I’m sure there’s an #R package that does it! I bet we can get it done the next week!

A few hours of googling later…

Eng: Oh, there’s nothing like that. Let’s download a dataset for it in .csv format from #Kaggle, try to code from there.

Next day…

Eng: Uh, there’s no readily available datasets, no out-of-the-box libraries. Let’s check Github!

After checking there

Eng: Guess, I’ll have to learn #Keras deeper and try to fix that.

After a few hours

M: Hey, how’s that pedestrian thing going?

Eng: Turns out the available model was good only for a demonstration! I’ll have to write my own in #TensorFlow

M: But you’ve told us that you’d be done by now.

Eng: This pre-implemented loss function is a mess! I will have to write my custom layers, loss function. so many things to study!

A course later…

Eng: Ok, it works, the simulations look pretty good!

Deploy eng: But it uses too much memory, we can’t use it in real time.

M: We have run out of funding and everyone is fired.

The moral of the story is that you can’t out google good fundamental education and relevant technical knowledge.