It's not a random number the College Board pulls out of their backside. At the end of the day, having an independent company running standardized tests across the nation in dozens of subjects (there's SAT and then there's SAT II) is really a good thing (TM).
Interviews are good, but they're not great. You can't have a good interview without good metrics at hand to qualify what you're seeing. There are geniuses that fail at social interaction and will buckle under the pressure of an interview at CalTech, MIT, Harvard, Stanford, etc. or Google, Microsoft, Facebook, & co. but you'd still want; just as there are interviewees that will ace any verbal interview with flying colors (yes, including technical questions) that you wouldn't.
At the end of the day: the SAT score, as it was intended to be used, is a datapoint. It's not the be-all end-all: it's an additional piece of information that should be used responsibly to arrive at a good conclusion whether or not to accept/hire an individual.
A good score on the SAT + a good score at an interview means that this person not only knows their stuff and can think on their feet, it also means they're hard-working and will do what it takes to do great (believe me, it takes MONTHS (or even years) of dedication to actually ace the PSAT or SAT). A good score on the SAT and poor interviewing means this person is a hard-worker but not particularly great at thinking on his or her feet. A good interview w/ poor SAT scores indicates someone that couldn't be bothered to study for even a month for an important event in their high school career - can you trust him/her to research thoroughly before taking on a big project?
The problem is not with the SAT. The problem is with the interviewers at colleges and workplaces around the country that treat it as the Holy Grail. Don't blame the College Board for this one.
Anyway, what would you have the SAT replaced with? Don't tell me you'd have colleges accept students based solely on the grades they achieved - in the United States we have some the highest educational disparity across the nation, and even across neighboring cities and suburbs. An A+ at one school may not even be a C at another. Teachers who's pay depends on the evaluation of their kids' academic performance are not necessarily the people you want giving them a number you'd bet the farm on. Just look at the recent Chicago teachers' strike: at the end of the day, pay and compensations aside, it rested on the fact that teachers don't want standardized tests to be used as metrics to determine the quality of their teaching, and they wanted to set the evaluations themselves.
The statistical balance of the scores on the PSAT/ACT/SAT/SATII is very well-studied and well-designed, just like with the USMLE medical board examinations and all other engineered standardized tests. There are easy questions, there are medium questions, there are hard questions, and there are flawed questions which can't be answered - before and after every examination, the questions are classified and then re-classified to avoid anomalous or unfair results.
Coming up with these questions, organizing the examinations, coming up with the statistical review methods, grading the exams, and getting the results to the universities is not cheap - I paid for all these exams when I was in highschool myself (and my family was lower middle class at best) and while I then considered them to be pretty damn expensive (and that's why I did my best to get the mark I needed the first time around and worked hard to make sure money was not being thrown away), in retrospect the pricing is very fair (although I hear USMLEs are outrageously expensive, but then again, so is everything associated with the medical industry).