Hacker News Reader

Comment by wwweston | Hacker News Reader

The news that there are apparently people with PhDs in Machine Learning who are finding the job market difficult enough they're applying for unrelated technical positions seems like it runs counter to the conventional narrative about high demand and trouble finding truly qualified scientists...

View on HN

To me it seems that the explanations in both cases are that humans are really illiquid. When I'm job-hunting, I want a job; I don't want to wait two months for a better one. (Part of this is my own desire; part of this is the social convention that companies who do manage to give me offers want answers quickly, and won't be interested in re-interviewing me two months later if I decline them.) When I have a job, it will take a significantly better job for me to risk the steadiness of my current job, the accretion of social capital and seniority/tenure, etc., and also to risk the unknown of whether management will be good at my new job, which is rather hard to determine even if you know people at the company. And I'm certainly not going to jump ship within about a year of starting a mildly acceptable job, and within much more than a year if I recently jumped ship after about a year.

So people are probably badly in need of for PhDs who are on the job market, but those PhDs are meanwhile badly in need of positions fairly quickly once they graduate, and if the market isn't big enough, both sides are going to reach theoretically-suboptimal outcomes because of lack of liquidity.

(Consequently, GP should not be afraid to hire those people. If they say "No, I want this job," chances are the majority of them will stay around for quite a while even if better opportunities open up.)

Or it means that supply is starting to meet demand. People interested in tech hear that an intellectually stimulating field like machine learning is highly lucrative and short in supply, what happends? People start entering PhD programs and out they come a few years later. That narrative is years old at this point. That original NY Times article on deep learning is now five years old, plenty of time for people to be finishing their degrees. Clearly it seems like the tides are shifting. And I don't think the supply is on the downturn at all, but demand definitely seems to have peaked since companies are starting to realize that none of the stuff that drove the hype (deep learning) is actually applicable to them personally.

Certainly true from what I've seen. The last company I was at was desperate to find a remotely qualified ML guy in the area (we were out of town), and three different startups I have talked to in the last two weeks have multiple openings for ML and asked me (who has no formal ML qual) whether I was interested.

So who knows. Maybe they don't want to move or it isn't sexy enough ML work for them.

>conventional narrative

That shit's been broke for employees since Agile was successfully marketed and they figured out how to reproduce it (blogs, news, and conventions lead to consultants). Same for Big Data, ML, AI, coding bootcamps, and DevOps.

Why? Apart from the big guys, who is actually doing deep research or highly technical work in Machine Learning? It is a niche industry.

I think there's some pretty high variance in the quality of machine learning PhDs - some professors are pretty upfront that they have a team of graduate students throwing algorithms at datasets and seeing what sticks. I can't imagine that approach trains great researchers.

Ya, something smells fishy here. On the other hand, I hear my PL PhD is great for working on ads at google.

Reply on news.ycombinator.com