Political arguments from extreme examples are extremely bad
Suppose you knew for a fact that your family was in imminent danger of being murdered in a plot masterminded by bin Laden or some other homicidal psycho? Suppose, further, that you knew a confession from him would save the lives of your family. Would you be willing to have the authorities use torture to obtain it?
or, if that doesn't make you sufficiently afraid,
Picture this: There's a guy in police custody who knows the location of a bomb that's set to explode in an elementary school building that holds 500 children. Threats and intimidation have failed to make him reveal the address of the school and the location of the device. You suspect that torture would loosen his tongue and save all those innocent lives. What would you do?
There are several flaws in this argument, one being that we rarely "know for a fact" events that have not happened yet, another being the dubious missing link in the logic here, that the information extracted would be reliable and thus help law enforcement prevent the danger rather than provide them with misinformation that distracts them from the real target. But there is a more basic problem with these kinds of arguments when applied to any argument over what any law should be.
The use of extreme examples preys upon the widespread ignorance in modern societies of how our legal systems work. It assumes that our legal system executes laws as if it were a robot: law goes in, facts go in, defendant goes in, from the legal code and the facts the robot logically concludes that the defendant violated law, defendant goes to jail, period. But that's not how our legal system works. Rather it is set up to assume, properly, that no language can describe just outcomes in all cases. It is set up so that when legal rules cause obvious injustices, those injustices can be avoided. In the United States and Commonwealth countries, at least three different groups can exercise discretion to prevent obvious injustice to technical lawbreakers: prosecutors, grand juries, and trial juries.
In the case of the torturer of an obvious terrorist who in fact saves many lives, it's highly likely that prosecutors would exercise their discretion to not prosecute. They'd get overwhelming political support in doing so. Even if they prosecuted, they would be unlikely to get a grand jury to indict. And even if a grand jury indicted the life-saving torturer, it's extremely unlikely that a trial jury would convict. Thus, despite violating the words of the law, our life-saving torturer would be hailed as a hero, get millions of dollars worth of book contracts, and face no serious risk of jail time. Strong anti-torture laws, with no exceptions written in to the language of the law, would prevent the gross injustice and inhumanity of torture in the vast majority of cases where officials would be tempted to torture -- cases where the dangers posed by the torture victim or his pals are more pedestrian or less clear than the extreme scenarios put forth by proponents of legalizing torture. In such cases the importance of not stooping to the barbarity of torture, and the extremely horrific possibility that torture might become standard law enforcement and military practice, should trump uncertain or everyday security considerations. Strong and absolutely worded anti-torture laws would not punish a torturer who clearly saved many lives in the extremely rare circumstances so often portrayed by these proponents.