I'd argue you want both work sample (for technical competence) and the biographical interview (to determine provenance, likely future behavior, etc.), and then combine with some "see how the person works in real time" situational/performance/behavioral interview.
The interviews which are worthless are: technical interviews conducted by non-technical people, meandering style unplanned interviews, and "language trivia" interviews conducted by the marginally competent.
Essentially, interviews should be formal, measurable, and then tracked against performance.
I'd put some basic pre-screening at the top of the funnel if you have too many candidates. My personal first stage filter right now is "referred by someone I know and trust", which is a very selective filter, but at a larger company you can use a less selective filter.
It's difficult to pull off because of possible discrimination claims, e.g., IQ tests are racially biased, but it can be done if you can demonstrate a business need.
I wonder if there are any examples of businesses that only hire people over a certain score on a general intelligence test?
I know that several other strategy consulting companies do use intelligence screen based on progressive matrix tests as part of their interview process, but my only experience there is European (but with American companies) and I don't know if they use it in the US as well.
But of course. And this is why hiring procedures should generally be based on work-sample tests rather than on reviews of resumes (which may just be a tissue of lies) or impressions after personal interviews (which may also be full of lies). Research shows that neither biographical data about job applicants (e.g., resumes) nor the performance of job applicants in interviews does as well as work-sample tests in finding workers who do good work.
Full references can be found for these facts in my FAQ post on hiring procedures,
http://news.ycombinator.com/item?id=4270768
last posted in complete form about a month ago here on HN.