I'd guess they have a very high drop-out rate.

Regardless, "machine learning" is a very broad field and honestly I have no idea what an "ML engineer" is doing if they are one. It can cover any of the following:

1. Cutting-edge academic research (do better on this test set)

2. Doing data analysis to identify prediction ability

3. Creatively thinking of useful features to evaluate.

4. Implementing data pipelines/logging to obtain the features needed for #3.

5. Production systems to evaluate/train ML systems. (multiple places in the stack).

Because the spectrum is so wide, if you are already an engineer, you can readily get into "ML" categories 3-5 and even 2. Andrew Ng's course is a valuable introduction and not that heavy of an investment -- I found that just with it (alongside my product and infra background), I could readily contribute to ML groups at my company.

>> 1. Cutting-edge academic research (do better on this test set)

It's interesting you put it this way. I think most machine learning researchers who aspire to do "cutting-edge" research would prefer to be the first one to do well on a new dataset, rather than push the needle forward by 0.005 on an old dataset that everyone else has already had a ball with. Or at the very least, they'd prefer to do significantly better than everyone else on that old dataset.

I bet you remember the names of the guys who pushed the accuracy on ImageNet up by ~11%, but not the names of the few thousand people who have since improved results by tiny little amounts.

> but as of right now, you'd be among "2,647,287 already enrolled!"

* Enrolling yourself in a free online course, does not, at all make you a ML expert. A very sizeable portion of those enrolled may not have gone further than the first chapter.

* The ML train (as in people actually knowing ML) is not, at all, crowded

* Even if the train was crowded, learning ML does not make you forget what you already know. Saying "I don't want to learn this skill because so many people already have it", just means there are that many people that now have one more skill than you (unless you decide to allocate this time to learning something else)

> learning ML does not make you forget what you already know

I often feel that learning something new takes such energy and mental rearrangement that it does crowd out what I already "know". The brain is a neural network whose weights are constantly being adjusted, just because you learned something at one point does not mean it is permanently there.

For example, I spent many years working in networking, but now that I've been out of that field for several years and working in embedded firmware and data, I would have to relearn much of what I "knew" in my old field in order to be professionally productive. And that's apart from the field itself advancing in ways I haven't kept up with.

It's like riding a reverse-steering bicycle. By learning something that's in direct competition with your existing neural structures, your neural structures change and you no longer "know" what you used to. Jump to 5:10 to see this person, who took 8 months to learn to ride a reverse bicycle, try to ride a straight bicycle: https://ed.ted.com/featured/bf2mRAfC

> but now that I've been out of that field for several years and working in embedded firmware and data, I would have to relearn much of what I "knew" in my old

This seems to me this is mostly caused by being out of the field for so long, not because you did something else while being out.

That seems like a distinction without a difference. The reason time erases skills is because you're always doing stuff, and over a longer period of time, you've done more stuff, which has reprogrammed your brain.

You have to be joking. Demand for ML skills, especially scaling in production, is off the charts.

Scaling ML in production doesn't take much ML skills though.

The bulk of industry ML work isn't actually suited to ML phds without engineering skills. I work in FAANG and this is a huge problem where the ML phds have poor communication with skilled engineers and a lack of engineering experience. They often even look down upon people who don't have fancy credentials. Unfortunately they just end up creating a money wasting disaster of a system.