While I love the number crunching ethos, and think these guys are giving it one hell of a go - I will go on the record and say that they will ultimately fail because they are operating under many false premises, a few being:
1. People do not change their minds.
By algorithmically matching based on answers to questions, you must assume that those answers are meaningful. And meaningful means CONSISTENT. And people are NOT consistent. Especially on complex issues; in fact there are whole professions (sales (insurance, real estate, car, etc) comes immediately to mind) where success is based on one's ability to get people to change their minds. (That example is offered as evidence, not proof - but proof probably isnt a complicated exercise)
2. Like Thinkers make Great Daters
Assuming you are even GETTING like thinkers, (which is tenuous at best) where is it written that like thinkers make great companions? You don't want to have absolutely nothing in common, (at least a common language is necessary) but it isn't linear and i would be surprised if there were any correlation let alone one that was non-asymptotic. Simply - more like thoughts does not entail more compatibility.
3. Looks aren't EVERYTHING
Just based on sheer observation - this is not the case. It is unfortunate, and it speaks badly about human nature, but most human beings make judgments based on appearances. And this "appearance crutch" is only exacerbated when it comes to looking for someone to lay with. How many single, male, bleeding heart liberals would take a swing at Sarah Palin if given an iota of an opportunity? Again - offered as evidence, not proof.