Once it's conceded that victims of torture will say anything, including the truth, only two hurdles remain: First, to establish a protocol that exploits this fact. Second, an ethical framework for establishing when and how much torture may be used.
Here's my protocol in a nutshell: Don't ask the detainee to tell the truth. Ask the detainee to make testable predictions. The cell's safe house is X. The email accounts to watch are Y. The shipments are placed in a storage locker at the bus station, marked with a Z. Etc. You can even start by asking the detainee to make predictions that you know have already tested true. The torture stops when the detainee starts making accurate predictions.
Torture should never be an open-ended dominance game. That's morally, and pragmatically repugnant. It should always be a sober, experimental exercise in testable predictions and behavior modification--if it's even used at all.
But I doubt there's anyone in this thread besides me who would be in favor of torture if they thought it worked. All this talk about whether or not torture works is a red herring. Nobody is honestly objecting on the grounds that it doesn't work.