Didn't Google admit recently that there was not a strong correlation between how well a developer did on their interview questions and how well they performed within the company?
It's because only successful candidates appear in the statistics. If Google let everyone in, you'd see a good correlation between how well developers do in the interviews and how well they do at Google. If you dump the 95% of people who don't perform well in the phone interview, then what you're left with is a bunch of high performers who performed well to different extents (where the purpose of the interview is not to gauge ability but to demonstrate minimum ability), where the extent to which they performed well is largely noise.
Either a positive or negative correlation would be bad in this case.
No such assumption made. If you have a positive correlation in your data, you're probably not placing enough weight on your interviews in your hiring process (i.e. you knew people weren't as good at interview but you still hired them). If you have a negative correlation you're similarly placing too much emphasis on interviews; you're likely hiring people who interview well over people who'd be better in the job.
You might still hire bad engineers, but the uncorrelated nature of your interview/job performance data won't help you explain that.
Considering how many google products have failed in the past and are how divided users are on the existing ones I would question whether google's culture is worth anything at all.
Seems to me that many of the big tech companies are losing the reality plot like coked up Hollywood starlets. I think we might see a paradigm shift in tech again soon including how recruitment is carried out.
Probably, but you'd expect that because the interviews only test part of what is required to be a good employee. It's just a in-precise filter - a bar they force you to jump over, which if you manage to do so pretty much proves you're not hopeless (and have enough dedication to learn the material). So amongst the guaranteed-non-hopeless people there is fairly random distribution of ability to perform well in the company.
No, what they found was that there was no correlation between doing well on brainteasers (like "How many golf balls can you fit into an airplane?") and job performance.
I am not sure if you read the article but they never claim that they have found a correlation between performance on technical interviews and ability on the job.
Yes, brainteasers are worthless and "structured behavioral" questions yield better informations but there is a gap between claiming that technical interviews are worthless and that.
Also, it is usually the duty of the author to support their claims with a direct reference to sources not the opposite.
