I have to agree. Especially given the very real possibility that your ML project won't be cutting edge research grade. At that point someone who doesn't have bias and is willing to search for a reasonable looking approximation to the problem and try a canned solution may actually be an optimal candidate.

Considering the number of problems that could be plugged into a random forest with good results, data proficiency seems more important than strong ML experience.

Depends heavily on the application once you get to more specialized domains.

I wish there was an easier way to label roles differently based on when you just need to throw X or Y model at some chunk of data and when more specialized modeling is required. Previously it was roughly delineated by "data science" vs "ML" roles but the recent AI thing has really messed with this.

>Just try to find someone with a burning ambition to learn. That seems like the key to get someone capable in the long run. If they point out something beyond Kaggle that makes you think, pay attention to that feeling — it means they’re in it for more than the money.

If you're teaching them, you shouldn't be paying them at the AI expert rate.

But we do this in software engineering all of the time, why is AI different?

Corporations love people with experience but they don't want to actually invest in creating those people. If nobody is supposed to hire people who have only taken classes or done tutorials, how do you actually get people who have that experience? Or are these guys expecting us to bootstrap our own PhD before they deign to speak to us?

This reminds me of when I started learning spark (back in the dinosaur days). It was considered this cutting edge 'advanced' technology that only the top tier of 10x engineers knew how to implement. The documentation was crap and there were not many tutorials so it took forever to learn.

These days people can get an excellent introductory class to spark and be just as good as I've ever been at it. I wouldn't call them 'charlatans' like the poster above did. It's just that the libraries used to implement spark have been abstracted and people learn it faster.

That's just how it goes in tech. Anyone who wants to learn is a treated like a poser. We over-index on academic credentials which are really not indicative of actual hands-on engineering ability.

PS. There are no AI/ML experts. There are LLM experts, prediction model experts, regression experts, image recognition experts.... If you are hiring a 'AI/ML expert', you have no idea what you are hiring.

If you can make do with generalist techies who can ramp up in a few weeks, you probably don’t need to be paying them $500k-$1M TCO. They’re just a new technician.

But that doesn’t mean that having people with actual research/depthful expertise aren’t essential and hard to find amongst the noise.

The person you responded to is talking about would-be technicians applying for researcher roles. That happens in tech booms and opens amazing doors for lucky smart people, but it’s also a huge PITA for hiring managers to deal with.

What you say is true but in terms of hiring and screening candidates it really depends.

One important aspect of all skills is to know the limitations and boundaries of said skill. It’s probably fine if somebody implemented ML on a trivial problem to learn and practice, but if they didn’t realize there could be better solutions in the first place and that ML isn’t a solution to everything, then it’s a big red flag for me.

Also, finding a good problem for a solution is also a handy skill, if one can’t figure out how to apply their skills to a real problem, then that does give slightly negative impressions.