I say it for the opposite reason of elitism (and I’ve spoken out many time against elitism in ML/AI).

As other posts note, there are many resources available for teaching the concepts. But they don’t teach the limits of AI, and the rise of MOOCs is setting novice programmers up for a shock when they encounter real world data that is not as nice as the Titanic dataset, and requires making smart decisions to handle the data, and handle it in a way that does not invalidate the results.

Many romanticize ML/AI as something that can solve any problem, which is a dangerous approach.