Physicist and mathematician Freeman Dyson on string theory:

But when I am at home at the Institute for Advanced Study in Princeton, I am surrounded by string theorists, and I sometimes listen to their conversations. Occasionally I understand a little of what they are saying. Three things are clear.  First, what they are doing is first-rate mathematics. The leading pure mathematicians, people like Michael Atiyah and Isadore Singer, love it. It has opened up a whole new branch of mathematics, with new ideas and new problems. Most remarkably,  it gave the mathematicians new methods to solve old problems that were previously unsolvable.  Second, the string theorists think of themselves as physicists rather than mathematicians. They believe that their theory describes something real in the physical world. And third, there is not yet any proof that the theory is relevant to physics.  The  theory is not yet testable by experiment. The theory remains in a world of its own, detached from the rest of physics. String theorists make strenuous efforts to deduce consequences of the theory that might be testable in the real world, so far without success.
. . .
Finally, I give you my own guess for the future of string theory. My guess is probably wrong. I have no illusion that I can predict the future. I tell [page-break] you my guess, just to give you something to think about. I consider it unlikely that string theory will turn out to be either totally successful or totally useless. By totally successful I mean that it is a complete theory of physics, explaining all the details of particles and their interactions. By totally useless I mean that it remains a beautiful piece of pure mathematics. My guess is that string theory will end somewhere between complete success and failure. I guess that it will be like the theory of Lie groups, which Sophus Lie created in the nineteenth century as a mathematical framework for classical physics. So long as physics remained classical, Lie groups remained a failure. They were a solution looking for a problem. But then, fifty years later, the quantum revolution transformed physics, and Lie algebras found their proper place. They became the key to understanding the central role of symmetries in the quantum world. I expect that fifty or a hundred years from now another revolution in physics will happen, introducing new concepts of which we now have no inkling, and the new concepts will give string theory a new meaning. After that, string theory will suddenly find its proper place in the universe, making testable statements about the real world. I warn you that this guess about the future is probably wrong. It has the virtue of being falsifiable, which according to Karl Popper is the hallmark of a scientific statement. It may be demolished tomorrow by some discovery coming out of the Large Hadron Collider in Geneva.” (page 221-222)

POSTSCRIPT (2012-12-27):  Physicist Jim Al-Khalili interviewed in The New Statesman (21 December 2012 – 3 January 2013, page 57):

Theoretical physics in the past hundred years has sometimes bordered on metaphysics and philosophy, especially when we come up with ideas that we can’t see a way of testing experimentally.   For me, science is empirical – it is about gathering evidence.  It’s debatable whether something like superstring theory, which is at the forefront of theoretical physics, is proper science because we still haven’t designed an experiment to test it.”

The link to metaphysics should come as no surprise, since all scientific investigations eventually end there, as Boulton argued.
Reference:
Freeman Dyson [2009]:  Birds and frogs.  Notices of the American Mathematical Society, 56 (2): 212-223, February 2009.   Available here.