I get that you are giving anecdotal examples. But your suggestions are very dangerous, more so for aspiring Data Scientists.

Drawing parallels between software programming and Data science is wrong. I think it is because both use code to express ideas, Software engineers think they can do data science too without having to know the maths/stats/ under the hood. Perhaps, a familiarity with code overshadowing unfamiliarity with under the hood science.

Your example of learning to write high level first and then dwelling on low level is not a good argument for Data science either.

"When you set up your password hashing, did you write a custom hashing algorithm? Did you spend weeks studying cryptography? Or did you just use bcrypt?"

No programming interview goes like "Hey code me a compiler" or "code me a hashing algorithm"

The reason is because these are all 'solved problems'. But Data Science is an explorative endeavor, in an essence a Data Scientist approaches an 'Unsolved problem'.

To solve an unsolved problem, one needs to have in arsenal 'under the hood science know-how' .

In a way, a data scientist is like the first person about to develop the first custom hashing algorithm. And to do that, yes, that person must study and know cryptography.

"Once you can build with ML, then learn the theory"

Quite frankly, there is absolutely no incentive to learn the theory here, if somehow a ML product is put together and some gullible person buys it.

Well if that gullible person comes back and complains about the product then one would go back and learn the theory . By then it would have been too late and destroyed the person's and company's reputation. Not to mention that 'gullible person' forming a belief that 'Data science is snake oil', which dents the entire industry .

I am afraid your suggestions could usher in AI winter sooner than expected !!

At the end of the day, it is all about skin in the game.

Would you pay for a ML product knowing that it was designed by a software engineer with absolutely no 'how things work under the hood' knowledge?

Would you buy and trust a ' AI powered Covid19 lung CT scanner' knowing that it was developed by a bunch of ppl who just randomly succeeded in tweaking some parameters and lowered some cost function?

If the answer to above questions is NO. Then I guess your suggestions are not right either.