Not really surprising at all. A huge amount of selection has already taken place by the time you get to the interview, and most of that selections is based either on grades or on other metrics that measure similar traits. The “low-GPA” individuals who get an interview are therefore very different from the general population of “low-GPA” individuals, and one expects there to be little or no correlation between GPA and success (if anything, I would expect negative correlation, as low-GPA individuals who get interviews will have something else going on that got them there).
This is almost exactly the same effect as the fact that SATs do not predict college GPA. SAT scores are used as one of the major factors for admission to colleges; having separated the students into cohorts based (partially) on SAT, it is completely expected that SAT scores have minimal correlation with grades assigned within each cohort.
The sorting continues in college; SAT scores are most predictive for the first year, before less-capable students switch out of the hard majors. My guess is that nearly all of the engineers Google hires score very highly, and in that sense the SAT is indeed predictive of success there (since you can take it in middle/high school), but they might have determined that discriminating on the high end between differences of <1 SD (about the validity between retakes) might not tell them much.
Actually they say they're giving less weight to them (historically Google has weighted them very heavily) which is significantly different from "test scores are worthless".
You've skipped over the original NYT article, which the qz.com article being linked to here quotes from. He does say they're worthless. Here's the complete quote:
One of the things we’ve seen from all our data crunching is that G.P.A.’s are worthless as a criteria for hiring, and test scores are worthless — no correlation at all except for brand-new college grads, where there’s a slight correlation
This chimes with my understanding and experience that Google only really use test scores and GPA's right now as a filter to manage the vast number of internship/entry-level applicants they get.
Makes sense, most companies only care about university grades for first jobs (because there's very little else to distinguish candidates at the stage).
I quit my job after 5 years as a programmer for my Masters. I must say, just working on one small class project with any candidate can help you find a world of a difference. I think its the experience and my personal interest that drives me towards worrying more about actually learning stuff instead of scoring grades. Getting a grade is more about identifying what a Professor expects and giving it to them. There are many students who slack off their effort in team projects. And, the worst part, the slackers spend a lot of time applying and preparing specifically for interviews, so as I see the ones who get the "best" jobs are generally the ones that I would never hire them if I was looking for candidates for my company.
This is almost exactly the same effect as the fact that SATs do not predict college GPA. SAT scores are used as one of the major factors for admission to colleges; having separated the students into cohorts based (partially) on SAT, it is completely expected that SAT scores have minimal correlation with grades assigned within each cohort.