X-Message-Number: 11507 From: Date: Sun, 4 Apr 1999 11:03:45 EDT Subject: intelligence, consciousness etc For Mike Perry and anyone else interested, let me try again, very briefly, to indicate why both extreme camps--strong AI people and meat chauvinists--are wrong. I begin by showing why both camps are, in part, correct. Searle's famous "Chinese Room" was intended to dramatize the fact that the ability to carry on an apparently reasonable conversation does not prove intelligence. His model was not the best, but he was absolutely right. A purely brute-force program of instructions, or automated Chinese Room ("Answer string Qn with string An") could be designed that would fool most of the people most of the time, if anyone wanted to bother investing the time and effort. Strong AI people note that, e.g., a computer that directs the operations of an oil refinery does not merely manipulate symbols that are empty of meaning. It does not just simulate oil refining; it actually causes oil to be refined. There is two-way interaction with the environment. Likewise, the most extreme AI people say that even a simple furnace thermostat thinks. Of course, it only has two thoughts--anthropomorphizing, either "It's too cold in here." or "It's not too cold in here." Yes, these really are thoughts, however primitive--if we agree that thought is any kind of goal-directed information processing. Along with many others, I prefer to go beyond that and require that "intelligence" include evolution or learning capacity as well as goal-directedness and problem-solving. Now let's see in what ways both camps (at least their typical spokesmen) are wrong. Searle says a computer only manipulates symbols with no semantic content, so knows nothing and understands nothing, regardless of its usefulness. But the most important communications signals in our brains are also just symbols. Yet they are correlated with physical conditions or events in the external world (or in other parts of our brains or bodies), and through two-way interaction or trial-and-error feedbacks we can exploit these symbols to build a meaningful representation of the world. It is also enlightening here to remember what might happen with attempted communication, say by radio or TV, with intelligent entities on a distant planet. We receive nothing but symbols--and yet ingenious plans have been devised whereby, without any Rosetta Stone, we could learn the aliens' language, just by examining the patterns of information. (Or we could teach them ours.) I believe this parallel is valid and powerfully convincing. Therefore ordinary computers do indeed have the potential of becoming intelligent, as defined above, and Searle's denial of this is wrong. But when we turn to consciousness and feeling, it is another matter. Here Searle et al are right and the strong AI people are wrong. This becomes clear when we take a hard look at the "hard problem" in consciousness--the anatomy/physiology of qualia--roughly, "feelings" or subjective conditions. Mike Perry has written: >Actually (and I may as well admit to being a strong-AI person, which I am), the thought that occurred to me is that, indeed a simulated stomach does not do the same thing as a real stomach (except relative to a simulated environment). I'll concede that it isn't really doing digestion! But, on the other hand, suppose we simulated a human brain. If our simulated brain is composing poetry or solving a math problem, then clearly the simulation is *really* doing these things and not just "simulating doing them". If our similated brain is experiencing consciousness, is this "real" consciousness or not? I.e. is consciousness more like digestion or more like problem solving, or neither? I would vote for it being "real" consciousness in any case. Subjective experiences are not tied to our physical world in the same way as chemical processes like digestion.< I think the last sentence is wrong. I don't want to repeat some previous long discussions, but it is reasonably clear to me that feeling is specifically physical and not just a pattern of information or the processing of information. It probably involves time binding, which is ruled out in Turing computers. I.e., the "self circuit" or "subjective circuit" involves something like a standing wave in the brain that includes an appreciably extended region both in space and time; feelings or experiences are modifications of that standing wave. Without that physical construct and REAL TIME correlations, there are no qualia, no consciousness, no LAWKI (Life As We Know It). A simulation of that wave would no more constitute feeling than a modem transmission of a scan of a photo of a rose would constitute a rose. In due course the experimentalists will verify or disprove this hypothesis (if I can call something so vague an "hypothesis"). This will not solve all the "philosophical" problems of criteria of survival--it may even make them harder--but it will prick a few bubbles. Robert Ettinger Cryonics Institute Immortalist Society http://www.cryonics.org Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=11507