[pfedor]

So do you believe that if you take a random sample of (eg) Google engineers, and a random sample of all software engineers, there will be no difference in ability between the two groups? Or do you think there would be a difference but attributable to something else than their hiring practices?

[tptacek]

I think conventional job interviews do allow you to target a broad range of competence, with a large margin of error, so if you were to compare Google employees to, say, State Farm line-of-business .NET application developers, sure, the Google engineers would outperform.

It's an interesting question whether a random sampling of Google engineers would outperform a random sampling of tech industry product developers from other companies. I don't know.

The bigger is, within the broad set of candidates that an interview process selects for, the conventional job interview is arbitrary bordering on capricious; it's inhumane for no business benefit.