THE INVERSE-TURING TEST
The Turing test measures a machines ability to be 'intelligent'. The test involves a person having a 'blind' text conversation with either man or machine. If the person is consistently unable to determine to whom they are talking, the machine is said to have passed.
To my knowledge, no machine has ever passed this test. In that sense this test can determine whether a 'being' is human. This is because the machine lacks sufficient 'human' intelligence.
Now consider the opposite question. Is it possible to form a test which can prove machinery? IE, humans will consistently fail this test. If we put a few restrictions on the human, I believe this to be so.
That test already exists, it is called the 'Diehard Test Suite'. It sounds like a simple test--can you generate independent and identically distributed random variables? At its simplest, could you generate a completely random series of 0's and 1's? The restriction: you can't image some sort of complicated algorithm in your head and you can't use information from your surroundings. That's it. And my hypothesis is that humans cannot do this because of the fundamental nature of how we think.
Wait, why the restriction on complicated algorithms? That's like saying your computer has to pass the Turing test, and by the way it isn't allowed to use language comprehension routines. Shouldn't the natural limitations of a person's ability to implement explicit mental algorithms be enough?
ReplyDeleteIf the sample size from the human was large enough I suppose a 'mental algorithm' of some sort would be alright.
ReplyDelete