X-Message-Number: 14937 Date: Thu, 16 Nov 2000 18:36:50 -0800 From: Lee Corbin <> Subject: Re: Simulating People and Animals Pat Clancy wrote >There's really no reason to think that the functioning >of the mind is an _algorithm_, which is what is required >to make it implementable as a computer program. According to the technical definition of "algorithm", algorithms halt! An "algorithm" is used in most texts to refer to a process that delivers a definite output upon the receipt of a definite input and then stops. In more recent books, some authors have loosened this usage, but most (e.g. Penrose) continue to say "algorithmic behavior" when they speak in general of what programs can do. A person or a program may not merely provide a mathematically certain output, but may instead tell you to go stuff it, or behave in some other unpredictable way. The behavior of the first robots who seem like people will be like this. An operating system is a good example of a program that technically isn't an algorithm (by this definition). An OS offers you a response for each possible input you provide it, but it's capable of much more. It has a long memory, so to speak. I'm sure that you can easily imagine an interactive robot which obeys commands and perhaps even gives some uppity back talk from time to time, just the way that operating systems seem to. Yet it still seems to you that in the course of millions of years of development, computer programs can never imitate humans in any way whatsoever? Please imagine a life-like robot that responds to all of the world's stimuli unpredictably. (Unpredictable, of course, unless you run a simulation of the same program.) The burden is still upon you to say why a tremendous amount of emergent behavior from an extremely complex set of programs cannot mimic animals or humans. You bring up Dreyfus' old book; believe me, many of us on the other side, e.g. Dennett, Hofstadter, and many many others do not find his arguments convincing. And as for Penrose, as wonderful as his books are (I am perpetually re-reading them), his views about mysterious goings on of microtubules and what not, have been panned by numerous people, e.g., Ralph Merkle. As Marvin Minsky said, to paraphrase, "Roger Penrose is so intelligent that there are only two things in the world that he does not in principle thoroughly understand. One is quantum mechanics, and the other is consciousness. So who can blame him for thinking that they must be somehow intimately related?" >So, even the prey-recognition or locomotion capabilities of a >"primitive" animal are still beyond the most sophisticated computer, >whereas chess grandmasters are in serious trouble. This is a big >clue that computers aren't the right things to implement minds. I never said that they were! Our claim is not that TMs are the _best_ way to implement artificial intelligence. (Perhaps whatever way that AI is eventually implemented will reduce to being equivalent, but that is a different argument.) >The key test is the so-called Turing test - the artificial >mind along with a set of real people are "behind a curtain" >answering questions, and if you can't pick out the AM then >it passes the test (i.e. it's "equivalent" to a real person). >I just don't think a computer will ever pass that test. Can you guess what would be the give-away difference? Again, after millions of years of development of programs by humans and other programs, what tell-tale clue could still be present? Must it write a sonnet (to use one of Turing's examples)? Why in principle can't a robot sing and dance? Even after just fifty years of development, they're very good at some things. Why is there some sort of vague barrier that forever prevents them from doing other things? Lee Corbin Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=14937