Hacker News new | ask | show | jobs
by drones 780 days ago
I'm a part of my university's competitive programming club, and the discovery of Devin being a fraud really upset a few club members (Many people who worked at Devin.AI were elite-level competitive programmers). It shattered the illusion that being a good competitive programmer would translate into being a good engineer overall. They were the victims of market hype.
1 comments
iLoveOncall 780 days ago
> It shattered the illusion that being a good competitive programmer would translate into being a good engineer overall

There was a study by Google years back that showed the exact contrary: https://catonmat.net/programming-competitions-work-performan...

link
itsdrewmiller 779 days ago
Not exactly - it just means google overvalued that experience in its hiring decisions. https://erikbern.com/2020/01/13/how-to-hire-smarter-than-the... explains the phenomenon ("Berkson's Paradox") well.
link
iLoveOncall 779 days ago
I really don't think this is applicable here.
link
itsdrewmiller 778 days ago
No offense but did you read the article? It also links to this one which is about this exact claim - https://erikbern.com/2015/04/07/norvigs-claim-that-programmi....
link
iLoveOncall 778 days ago
Yes I did, I even read other sources about the paradox to make sure everything was clear in my mind.

The whole argument relies on the fact that "being good at programming contests" is a factor that was considered on interviews (and even given too much weight in the decision). There is absolutely no hint of that at all, and having been on the hiring side in FAANGs I can safely say this is not the case.

There is no paradox here because the study is not done on a pool of candidates that are pre-filtered based on this specific parameter, only on what it correlates (being good at LeetCode, which makes you pass the interviews).

It is funny that in the second article you linked the author says "My point here is that you can tweak these variables and end up seeing correlations with pretty much any value.", because this is exactly what he does. He manipulates the problem until it turns into a Berkson's paradox.

link
sujayk_33 778 days ago
It's like when a person goes too deep into CP, he just overfits in it and might just not come out of that phase, where you are given the problem and you are supposed to get the pre-defined solution.

It's not the case in real-life, the main problem is to detect the problem and then find the optimal solution. So unless you blend into it, get the test of CP, develop logic and get outside, build projects and keep moving, the field is vast, don't keep knocking same door everyday.

link