Haskell vs Java job applicants

That's not necessarily true; for things like designing proprietary languages themselves optimized for financial risk modeling, yes, you better have a CS Phd. But it's definitely not the case that everyone who touches functional programming on Wall Street has a Phd, and if you interview with the right group within an investment bank, the technical interviewer might very well be impressed with your Haskell experience, especially if you're going to be working closely with quants (provided he knows what it is; there's a really massive range in CS know-how within investment banks, but that's another story)

No one's saying that (I think). The claim being made is that someone who wants a job programming in Haskell is likely to be someone who's interested in learning new things, not that someone who has learned some Haskell is likely to be a good programmer.

Still, the thing linked is distinctly light on evidence (not a fault -- it's only an email on a Haskell mailing list) and I'm not quite sure what it's doing on the front page of HN; it's not exactly news that people wanting jobs programming in languages that are (1) difficult to learn, (2) interesting and (3) unusual, tend to be people who are interested in learning new stuff and keeping their brains active.

Being able to write a "hello, world" in Haskell certainly doesn't make you a good programmer. However, I would say that anyone who has already written a sizable Haskell program (5K+ LOC) is almost certainly a good programmer.

The reason is that it's more work to grok Haskell than almost any other existing language, and the job market for Haskell hackers is practically nonexistent. You have to really love programming to become a proficient Haskell hacker.

Surely numbers are significant here. How many universities teach Haskell, compared to the number who teach Java? Of those that do teach Haskell, how many of them teach it to a level that someone taking that degree would easily be able to extend that knowledge to a level where they could get a job programming Haskell on a day-to-day basis?

It's also a question of perception. The programmers one wouldn't want to hire, according to the metric under discussion, are those who learn a single popular language just to get a job. Is Haskell generally perceived as a language where there are a lot of jobs floating around? If not, then presumably most people learning it will be learning it because they're interested, because they want to improve as programmers, and so on.

I would say that there are many more that don't even offer Haskell at the advanced level, or at best, offer one or two classes in the entire curriculum that use it.

Which schools are that?

What if its for a blub job, say, maintaining 20-year-old billing systems? There's no value proposition in doing a rewrite, and the Haskell guy would get bored silly, while the blub guy would be content, and plug away, doing solid work for years. He makes a career out of it. And he was cheap. Everyone wins. It happens all the time. It's perfectly reasonable to hire a blub programmer. Whether it's reasonable to be a blub programmer is a different story. :)

The value proposition wasn't in simply hiring a "Blub Programmer," it was in identifying individuals more likely to be intrinsically motivated to improve their mastery of programming.

The majority of Java hackers picked it up (directly or indirectly) due to its market presence. Myself included.

The majority of Haskell, Clojure, and Blub++ programmers are hacking away out of personal interest.

Edit: On review, cesqui pointed out downthread that pg already covered this premise in http://www.paulgraham.com/pypar.html

The purpose of a hiring filter is to reduce the false positives. Your discovery of false negatives doesn't say much about the filter quality in this respect.

If you could take random samples from the group of people who would apply to Haskell jobs and from people who would apply to Java jobs and come up with a measure of motivation/willingness-to-learn, I believe there would be some statistically significant things to say. Samples from two mainstream (and in my opinion, trashy) languages like C++ and VB might be harder to differentiate.

As a side note, I've noticed that companies that can afford a large amount of false negatives tend to have horrible interviewing practices. Google is in the top two of the worst interviewing experiences I've ever had. When everybody salivates over your company, you can afford use a quick and dirty filter that hones in on any one arbitrary characteristic. As a matter of fact, one could possibly argue that these lazy filters do better than more accurate filters in the same way that Google argues that more data is better than superior AI. It's stupid for people to look to companies like these for exemplary models of interviewing.