Larry J. Crockett, «The Turing Test and the Frame Problem: Ai's Mistaken Understanding of Intelligence»

×

Larry J. Crockett, «The Turing Test and the Frame Problem: Ai's Mistaken Understanding of Intelligence»

Close

Part of what it means to be a researcher is to identify what appears to be a relationship that others either have not noticed or have not fully appreciated. Both the Turing test and the frame problem have been significant items of discussion for more than 20 years in the philosophy of artificial intelligence and the philosophy of mind, but there has been little effort during that time, as I read the literature, to distill how the frame problem bears on the Turing test. I believe that there is an important relationship and that this essay articulates that relationship. Of course, I must leave it to the reader to judge whether I have argued convincingly that such a relationship exists.
In a larger sense this essay is also an attempt to explain why there has been less progress in artificial intelligence research than AI proponents would have believed likely 25 to 30 years ago. As a first pass, the difficulty of the frame problem would account for some of the lack of progress. One could take this to mean that the progress will simply be much slower than originally anticipated because the problems are much more complex. An alternate interpretation, however, is that the research paradigm itself either is destined to be less productive than we might have hoped or must undergo radical change. In general terms, the view advanced here is that the future of AI depends in large part on whether the frame problem will fall to computational techniques. If it turns out that the frame problem is computationally intractable, if there is no way to solve it computationally by means of a program operating on formally defined constituents, which is AI's understanding of intelligence, then I suspect an ncreasing number of writers will reach the conclusion that AI embodies a fundamental misunderstanding of ntelligence. In fact, that is the view I tentatively advance here.
In short, I defend the view that the Turing test is a good test for intelligence but that no computer will be able to pass the test unless there is a solution to the frame problem. Against those, such as Gunderson and Searle, who say that passage of the test would not necessarily signify that a machine was thinking, I argue that a computer passing the test would have to possess a solution to the frame problem and that such a solution should convince us that thinking is in fact occurring. Against those, such as Dennett and Turing, who imagine that passage of the test is just a matter of time, I argue that it is not at all clear that we will be able to solve the frame problem, and, as a result, it is not at all dear that a computer will ever pass the Turing test. In fact, I think the prospects for a solution to the frame problem are not good.