IRCS Conference Room
Department of Computer Science
Does the Turing Test need fixing?
Every once in a while, there is a small flurry of activity around the question of fixing perceived problems with the Turing Test, Alan Turing's proposal for a sufficient condition for attributing thinking to a computer. There is such a flurry going on right now, with articles in the popular press with titles like "Forget the Turing Test" and "What Comes After the Turing Test?" I will rehearse, yet again, why there is really nothing wrong with the Turing Test in the role Turing intended for it, taking the opportunity to describe a new result countering the best (and really the only) argument against the Turing Test as a criterion for thinking. Then, I will clarify where the kernel of truth in the current objections are. The issue has special implications as the field of AI is in the midst of designing new tests of research progress.