dmb said: ...Hubert Dreyfus says that artificial intelligence will never work. He teaches Heidegger at Berkeley now, but started out in the sciences at MIT. He has the very tough job of trying to explain to the IT community that they are working with certain metaphysical assumptions that lead them to error. ... They're typical SOM scientists.
Ian replied: Agreed - AI "will never work" until it is realised that life has to evolve before intelligence .... and then it's not artificial any more, simply real. Of course plenty of "IT" people do see that ... the enlightened ones ... so your generalization is a bit swingeing. Krimel replied likewise: I would say it is way to early do discount AI. Moore's law is still ticking away and machine capacities continue to accelerate. It is impossible to say what capabilities inorganic intelligences will have in 20 or 30 years much less 100 years. ...But it is ridiculous to assume that philosophical analysis can declare anything to be technologically impossible. dmb says: I don't know that Dreyfus makes a biological argument. I mean, its great if the "enlightened" ones working on artificial intelligence understand how cognition evolved but I don't think that necessarily has anything to do with a shift in metaphysical assumptions. Please feel free to present a counter-example or otherwise correct me here, but I got the impression that there is a very limited number of people working in this area and that the whole approach was fundamentally flawed. I got the impression that Dreyfus wasn't trying to discourage them so much as simply explain why it hasn't, doesn't and can't ever work. And the flaw is such that even wildly increased "machine capacities" won't make any difference. Like I said, Krimel, you could input every fact in the world and it still wouldn't matter. Its clear that you think the whole suggestion is "ridiculous" but think about it for a moment. Its not even about the limits of engineering or the possibilities of technology so much as the limits of the assumptions upon which science has operated for several hundred years. They're basically trying to create a machine version of the subjective self in an objective world but the failure to do so is not a technical failure, per se. The problem is extending SOM too far, like trying to use Newtonian physics to explain the subatomic realm or relativity. The limits of those assumptions are exposed by the efforts in AI in the same way. The standard conceptions begin to fail in these areas. One of the ways to get at it is by way of language. The developments in the understanding of language over the last century or so have led people to say things like, "we are suspended in language". You were involved in the recent thread on that topic, eh Krimel? Here I would construe this as a shift from a correspondence theory of language, where we name things seen and express thoughts thunk, to a structural theory of language. It started with the semiotics of Saussure and Peirce about 100 years ago, became structuralism, then post-structuralism (which means MORE structuralism rather than AFTER structuralism) and these days deconstructionism is the thing. This way of thinking spread into areas outside of linguistics so that we can rightly count Piaget and Kuhn as structuralists. The idea, basically, is that words don't have meaning in their relation to the thoughts and things referenced so much as they have meaning in relation to each other within the total system, the whole context. You'll often hear people talk about language in terms of a "web" of beliefs, for example. You'll hear developmental psychologists talk about stages of growth in terms of a shift to a whole new gestalt or philosophers of science talk about paradigm shift. In all these cases, there isn't just more and more of the same but a shift in the whole structure of understanding. Heidegger had a very muscular version of this in his thinking. He said, "language is the house of being" and "being is that on the basis of which all things are intelligible". Its a strange way of talking, but I think its just a grandiose version of those cultural glasses Pirsig describes. I think our way of being is the web of beliefs we all acquire as we acquire language and they're both saying that there is more than one way of being. There are any number of ways of being and it shifts from time to time and differs from culture to culture. Language, then, more or less dictates how we see the world, or rather it determines the shape of our world. Thus language is the house of being. Language is the world we live in, not a reflection of the world we life in, see? The AI guys think robots live in our world, so to speak, and don't realize that our world is not THEE world. They want to have the intelligibility without the being upon which it is based. Am I making sense? dmb said: And there will never be a transporter like they have on Star Trek either. Its impossible for the same reason. Krimel replied: The idea behind the transporter is that the machine can read, store and transmit highly complex patterns. The holodecks and replicators operate of a similar principle in that they can reproduce stored patterns. It would seem to me this emphasis on pattern recognition and replication it right in tune with the MoQ. dmb replies: Yes, every viewer over 12 years old knows how it works. Thanks all the same, Dr. Science. The transporter idea depends on the assumption that a person is identical to their physical structure, that a person can be taken apart, shipped and re-assembled like a machine. And somehow, the person's consciousness could travel in the stream of de-patterned atoms. That's the hard part. In that sense, the holodecks and food replicators are far more plausible. If I ever had a holodeck, I'd spend some virtual time with Kate Beckinsale. You know, because she's really, really into Heidegger. Thanks, dmb _________________________________________________________________ E-mail for the greater good. Join the i’m Initiative from Microsoft. http://im.live.com/Messenger/IM/Join/Default.aspx?source=EML_WL_ GreaterGood Moq_Discuss mailing list Listinfo, Unsubscribing etc. http://lists.moqtalk.org/listinfo.cgi/moq_discuss-moqtalk.org Archives: http://lists.moqtalk.org/pipermail/moq_discuss-moqtalk.org/ http://moq.org.uk/pipermail/moq_discuss_archive/
