GODDEN@gmr.COM (01/15/88)
I would like to recommend to the readers of ailist that they take a look at the book >Engines of Creation< by K. Eric Drexler of MIT. He presents in layman's terms the basics of nanotechnology, which is the emerging field of molecular-sized machines, including computers. (In the notes are references to technical works.) Of particular interest to AI folks is the chapter on AI and nanocomputers. Let me just relate one item to give you a hint of what it's about. Drexler makes the fascinating claim (no doubt many will vehemently disagree) that to create a true artificial intelligence it is not necessary to first understand intelligence. All one has to do is simulate the brain, which can be done given nanotechnology. He suggests that a complete hardware simulation of the brain can be done, synapse-for- synapse and dendrite-for-dendrite, in the space of one cubic centimeter (this figure is backed up in the notes). Such a machine could then just be allowed to run and should be able to accomplish a man-year of work in ten seconds. The unstated assumption is that a computer that is isomorphic to the human brain will ipso facto be intelligent, and presumably will be able to construct its own 'mental' models once power is supplied. No need to supply it with software. (I may be misinter- preting the book on this point.) Interesting reading in any case. He even predicts (!) in chapter one that the initial nanomachines will be with us in ten to fifty years. Forward is by Minsky. In paperback. -Kurt Godden godden@gmr.com
dwt@EECS.UMICH.EDU (David West) (01/20/88)
In article <8801180618.AA08132@ucbvax.Berkeley.EDU> GODDEN@gmr.COM writes: > [...] the book >Engines of Creation< by K. Eric Drexler of MIT. [...] >it is not necessary to first understand intelligence. All one has to do is >simulate the brain [...] a complete hardware simulation of the brain can be >done [...] in the space of one cubic centimeter [...] h a machine could then >just be allowed to run and should be able to accomplish a man-year of >work in ten seconds. The breathtaking simplicity of the idea is awesome. Of course, some technological advances will be necessary for its realization, but note that to attain them, it is not necessary to understand technology ... all one has to do is simulate its development. A complete hardware simulation of the U.S. technological enterprise can be done in the space of one cubic meter (see appendix A) ... such a machine could then just be allowed to run, and should be able to accomplish a century of progress in one hour.
lewis@smu.UUCP (Eve Lewis) (01/30/88)
Re: Godden's review of >Engines of Creation< by K. Eric Drexler > Drexler makes the fascinating claim (no doubt many will vehemently > disagree) that to create a true artificial intelligence it is not > necessary to first understand intelligence. All one has to do is > simulate the brain, which can be done given nanotechnology. He > suggests that a complete hardware simulation of the brain can be > done, synapse-for-synapse and dendrite-for-dendrite, in the space > of one cubic centimeter (this figure is backed up in the notes). > Such a machine could then just be allowed to run and should be > able to accomplish a man-year of work in ten seconds. The unstated > assumption is that a computer that is isomorphic to the human > brain will ipso facto be intelligent, and presumably will be able > to construct its own 'mental' models once power is supplied. No > need to supply it with software. Perhaps we can suggest an approach to dovetail with, and enhance, Drexler's. One supposes that the structure of the brain is alright, as far as it goes. It's even been referred to as "Nature's Masterpiece." Nonetheless, it's important to understand that the human brain per se, is a product of the structural genes, to wit: the respected exons. I say the following: 1) Get an advance copy of that gene map of the entire human genome, a project now underway. (Be apprised that the human genome is the nanocomputer to end all nanocomputers, and anything else that comes along can only be second-best.) 2) Discard the real junk, the exon sequences. 3) Retain the alleged "junk," the intron sequences. Design a program to look for matches between two stores of complex data. Feed all the intron "junk" (one store of complex data) into the program. Accumu- late from all cultures, but particularly from Western "civ," a plethora of human thought fossils, i.e., Edward O. Wilson's cultur- gens (the second store of complex data), and feed the stuff into the program. Set the program going, to look for "matches," and voila! There you have it. A functional facsimile of the code underlying human thought processes. > The unstated assumption is that a computer that is isomorphic to > the human brain will ipso facto be intelligent, and presumably > will be able to construct its own 'mental' models once power is > supplied. No need to supply it with software. 4) The structure of the human brain is an excellent starting point, and I wouldn't neglect it entirely, "form following function," and all that sort of thing. Nonetheless, when push comes to shove, molecular biology is really where it's at. I mean, there are degrees of fidelity, when it comes to isomorphism. Besides, what A.I. really needs, is isofunctionalism. For example: One could construct a life-sized metal and plastic model of the human brain, even one good enough, in terms of isomorphism, for teaching purposes in a neuroanatomy class. Even one that could be taken apart, and put back together, with the nuclei properly enscon- sed, and the tracts properly aligned. If one then put clock innards inside, and supplied electrical power, would it "think"? We know that it would not. Indeed, if we imbedded a small clock face, with hands and numerals, in Broca's speech area, it would give us excellent time, just like any old Westclox!!! If we put in one of those talking mechanisms, Broca's speech area would TELL us the time. But our poor isomorphic brain, here, would not come up with Einstein's Theory of Relativity, or anything like it. And that's really what we're after, isn't it? 5) Now, what kind of a mind would we really like in our A.I. act-alike? Do we want the mind of a conformist? Or do we want the mind of a meshuggener? No, on both counts. We want the mind of a quality Zeitgeist smasher. So we have to discern the in vivo give-and-take of the repressors and enhancers, as well as the coding sequences themselves, and work that into any program, along with the base pair sequences that determine the culturgens. ("This one's no good," "That one's terrific," "Maybe we can use that one another time," etc.) We may even be able to avail ourselves of binary simplicity by equating the purine bases with 1, and the pyrimidine bases with 0. The entire thing is degenerate, anyway, and could do with some streamlining. 6) Some closing comments: Don't get hung up on morphology; it's a Linnaean trap, set for the sentimental. There isn't creature, a physiological structure, neurons included, or a neurotransmitter, that doesn't owe its life to several strings of base pairs. Really, it isn't the brain that thinks, or the neuron(s) that think; it's the differentially-expressed genome in the neuron(s) that thinks. In regard to the disdained pseudogenes, referred to as "junk" DNA, and believed to be "silent," which like Rodney Dangerfield, "Don't get no respect," perhaps like Alexander Fleming's penicillin mold, they could be the "bluebird of artificial/natural intelligence, right in our own DNA," and so, worthy of attention. It holds true for natural intelligence, artificial intelligence, and impressive isomorphs: "You can lead a 'mind' to information, but you can't make it think." And finally, hopefully there are no male chauvinists among the readers of, and contributors to, this net. If so, let me warn you that any aspirant to the Holy Grail, i.e., the comprehension of in- ternal rep, is doomed from the start if he overlooks the "junk" DNA in the maternally-inherited mitochondrial genome. - Eve Lewis
dwt@EECS.UMICH.EDU (David West) (02/02/88)
Let me be quite clear: in my earlier posting I intended to ridicule neither Eric Drexler nor the idea of molecular machinery. I *did* intend to ridicule the idea that meaningful simulation is possible in the absence of sufficient knowledge and understanding of the system one is allegedly simulating. -David West
craig@unicus.UUCP (Craig D. Hubley) (02/02/88)
>In article <8801180618.AA08132@ucbvax.Berkeley.EDU> GODDEN@gmr.COM writes: >> [...] the book >Engines of Creation< by K. Eric Drexler of MIT. [...] >>it is not necessary to first understand intelligence. All one has to do is >>simulate the brain [...] a complete hardware simulation of the brain can be >>done [...] in the space of one cubic centimeter [...] h a machine could then >>just be allowed to run and should be able to accomplish a man-year of >>work in ten seconds. > >The breathtaking simplicity of the idea is awesome. Of course, some >technological advances will be necessary for its realization, but note that >to attain them, it is not necessary to understand technology ... all one has >to do is simulate its development. A complete hardware simulation of the >U.S. technological enterprise can be done in the space of one cubic meter >(see appendix A) ... such a machine could then just be allowed to run, and >should be able to accomplish a century of progress in one hour. The bounding factor on progress thus becomes imagination. One could argue that this has always been the case anyway. The human race's primary occupation would then become dreaming up strange ideas for it's computers to chew on, prove/disprove, design and build. This seems almost natural, since our primary occupation has changed over the past three hundred years from manual labour through operating machines to moving information around. The so-called `Third Wave' of information technologies has only recently (within the last ten years) been widely recognized as such. It seems that you only sees the waves as they wash over you. Drexler's arguments, for those of you who haven't read the book, are broadly-based and in places expressionistic, though his appendices spell out in some detail his reasoning, and several chapters contain a sort of `question and answer' section where what must be the most commonly asked skeptical questions are themselves addressed. This is an intriguing technique of `compressing discourse' that more controversial books might benefit from, that is, an explicit answer to questions that otherwise would nag and bias the reader. If the answers are unsatisfactory, so be it. At least they are there to refute. I think it noteworthy that I've seen Drexler's name and book mentioned in several electronic and a few conversational forums, and not once did I ever hear an argument that he didn't explicitly address in his book. Nor have I heard a credible refutation of any of his points. On the contrary, I have heard nothing but enthusiastic recommendation of the book from those who've read it and receptivity to the ideas from individuals qualified in the specific fields concerned, from computing to molecular biology to business. Some of these individuals were very much skeptics at heart. I guess I won't be comfortable until I hear somebody *flame* the damn book! After all, it's annoying to have to just wait around wondering if I'll ever be able to solve problems just by thinking of them, live forever (barring accidents) in whatever environment I choose, and live in a body fortified by a truly formidable defense and immune system. If it's coming soon, I don't see much point in doing anything other than working on it, for those of us in technical fields. To solve the pollution, resource, food problems at once as a side effect! I'm afraid that reading this book puts truly big ideas into one's head. Don't read it unless your head is big enough to contain them. :-) And won't *someone* please flame the book!!! Craig Hubley, Unicus Corporation, Toronto, Ont. craig@Unicus.COM (Internet) {uunet!mnetor, utzoo!utcsri}!unicus!craig (dumb uucp) mnetor!unicus!craig@uunet.uu.net (dumb arpa)
mmt@dciem.UUCP (Martin Taylor) (02/11/88)
> ... such a machine could then just be allowed to run, and >should be able to accomplish a century of progress in one hour. I think we already do that, and have over the course of evolution managed such a speedup several times. No reason why it shouldn't happen again. The building of structure (information, organization ...) is recursive. The more you have, the easier it is to get more (a bit like money, come to think of it, and for much the same reason). BUT...humans will not participate in this greatly augmented progress, any more than green algae participate in human progress (except perhaps to be damaged by the side- effects, analogous to pollution and destruction of habitat with which we destroy those that have not shared our "progress"). -- Martin Taylor ...uunet!{mnetor|utzoo}!dciem!mmt mmt@zorac.arpa Magic is just advanced technology ... so is intelligence. Before computers, the ability to do arithmetic was proof of intelligence. What proves intelligence now? Obviously, it is what we can do that computers can't.