> Artificial intelligence engineers earn anywhere from 8% to 12.5% more than their non-AI counterparts,
What are AI engineers, and how'd they find their counterparts?
My impression is that a typical user of Web frameworks is going to have a much harder time learning enough about ML to do much more than make Web API calls to an LLM service and cargo-cult some prompts few actually understand.
I work for a big tech co and I have been involved in defining our hiring process for ML Engineers for the last couple of years.
Right now the loop to hire ML Engineers is the same as Software Engineers, swapping one coding round for one ML round. The content of this ML round is highly debated, although most interviewers seem to do a ML knowledge quiz type round (these are often conducted by data or research scientists).
From my experience we mostly get two kind of candidates:
a) People from academia, with masters or PhDs in ML.
b) developers who built a ML model in the last job and are trying to transition by jumping companies.
So far finding people who excel at both things have been really really hard.
Then there is the whole thing about whether we should test people for ML expertise, and the best way to do it. My personal opinion is that I prefer to hire strong engineers and then have them learn on the job. A general pushback I receive from hiring managers about this is that they have no ML knowledge whatsoever in their team, and they need to hire experts to get going. Which then raises the fun chicken-and-egg problem of how do you hire experts in a field that you know nothing about.
TL;DR What are AI engineers? My take is no one has a fucking clue.
If you need actual research and doing something innovative (openai, midjourney) get the academia people.
If you need to train models to deliver value to business, definitely go with the strong engineers. They can learn the ropes, there is plenty of material on the internet and their code will be usable and maintainable.
(I'm working with code written by academics and I won't recommend it)
We already have a role, research scientist, which would fit your first description.
Most of what people in the ML Engineer role would do in my company fit your second description.
But of course not everyone in the company, let alone the industry can agree on what are the expectations of a "ML Engineer", hence the shitshow we find ourselves in.
> We asked Oracle, the purveyor of databases, what it is doing in the world of gigantic data, but were told "whilst we have reams of information on big data, this is not an area in which we are experts".
If the pay difference is only 8% then on average we're probably only talking about a superficial experience difference. $800k, on the other hand, is probably actual researchers who understand the underlying math and aren't just gluing API calls together.
What are AI engineers, and how'd they find their counterparts?
My impression is that a typical user of Web frameworks is going to have a much harder time learning enough about ML to do much more than make Web API calls to an LLM service and cargo-cult some prompts few actually understand.