I can't help but notice your username, your critique, and then think of this: http://www.telegraph.co.uk/history/world-war-two/11151478/Co... (WW2 Bletchley Park using a crossword puzzle as a recruiting tool).
Those measures were better than nothing, but not better than a good measure.
Edit, because I want to amplify on this a bit.
In the absence of any other measure, I'm willing to believe that someone who's good at brain teasers is somewhat more likely to be a good programmer than someone who isn't (though the correlation isn't going to be perfect, by any means).
The thing is that I can also believe that someone who's good at brain teasers is more likely to be a good surgeon or mechanic than someone who isn't.
Yet we don't use this type of question when we're interviewing surgeons or mechanics. Why not?
Edit: I don't mean to suggest that I think these kinds of questions are good for hiring software engineers, just that your argument for why that isn't the case may have some holes in it.