I've long said that we shouldn't waste time arguing that torture doesn't work. For one thing, the evidence for those arguments seems empirically shaky, especially since many people employing them insist on arguing that torture basically never works, rather than that it doesn't work very often and therefore has a bad cost-benefit ratio. For another, arguing that something doesn't work isn't necessarily an argument for not doing it--it could just as easily be an argument for improving our technique. And if advances in brain scanning research let us develop a reliable lie detector, as seems possible in the relatively near future, then torture will work very, very well.
If that happens, we're in a nasty spot. Most people who make this argument do not, in fact, care whether torture works. They would still be every bit as much against it if waterboarding worked perfectly. Yet when they argue about whether torture works, they're conceding that torture's effectiveness is relevant to the question of whether or not we should engage in it. That implicitly means that if torture becomes nearly perfectly effective, they should change their minds--otherwise, it's not a relevant criteria. So if we get that lie detector, they have to explain why we still shouldn't use this very valuable interrogation method--or confess that they're basically opportunists who will say anything that might advance the case. This will make it somewhat harder to convince people to listen to their other, better arguments.