From extropians-request@gnu.ai.mit.edu Mon May 31 03:54:41 1993 Return-Path: Received: from usc.edu by chaph.usc.edu (4.1/SMI-4.1+ucs-3.0) id AA06566; Mon, 31 May 93 03:54:39 PDT Errors-To: Extropians-Request@gnu.ai.mit.edu Received: from wookumz.gnu.ai.mit.edu by usc.edu (4.1/SMI-3.0DEV3-USC+3.1) id AA01476; Mon, 31 May 93 03:54:36 PDT Errors-To: Extropians-Request@gnu.ai.mit.edu Received: by wookumz.gnu.ai.mit.edu (5.65/4.0) id ; Mon, 31 May 93 06:47:03 -0400 Message-Id: <9305311047.AA29878@wookumz.gnu.ai.mit.edu> To: ExI-Daily@gnu.ai.mit.edu Date: Mon, 31 May 93 06:46:40 -0400 X-Original-Message-Id: <9305311046.AA29871@wookumz.gnu.ai.mit.edu> X-Original-To: Extropians@gnu.ai.mit.edu From: Extropians-Request@gnu.ai.mit.edu Subject: Extropians Digest V93 #0291 X-Extropian-Date: Remailed on May 31, 373 P.N.O. [10:47:02 UTC] Reply-To: Extropians@gnu.ai.mit.edu Errors-To: Extropians-Request@gnu.ai.mit.edu Status: OR Extropians Digest Mon, 31 May 93 Volume 93 : Issue 0291 Today's Topics: Chaitin, Holland, & AI [2 msgs] ECON/SPACE: Sell the Universe Whole [2 msgs] Extropian Currency [1 msgs] Extropian Death Rituals [1 msgs] Fetishism/Speech Recognition/Extropian Business Bureau [1 msgs] HEALTH:Kenneth Seaton & Albumen? [1 msgs] MEDIA: ExI: BBC Radio 4, 0905 2 June 93 [1 msgs] META: Split the list [1 msgs] New Brain Test [1 msgs] New Brain Test: "Einstein einstein.... Penrose!" [3 msgs] PGP: A paranoid scenario [1 msgs] SPACE: Fermi paradox, etc. -- probably dumb question [1 msgs] Speech recognition [1 msgs] TECH:Cold feet on cold fusion [1 msgs] fetishism? [2 msgs] Administrivia: This is the digested version of the Extropian mailing list. Please remember that this list is private; messages must not be forwarded without their author's permission. To send mail to the list/digest, address your posts to: extropians@gnu.ai.mit.edu To send add/drop requests for this digest, address your post to: exi-daily-request@gnu.ai.mit.edu To make a formal complaint or an administrative request, address your posts to: extropians-request@gnu.ai.mit.edu If your mail reader is operating correctly, replies to this message will be automatically addressed to the entire list [extropians@gnu.ai.mit.edu] - please avoid long quotes! The Extropian mailing list is brought to you by the Extropy Institute, through hardware, generously provided, by the Free Software Foundation - neither is responsible for its content. Forward, Onward, Outward - Harry Shapiro (habs) List Administrator. Approximate Size: 53045 bytes. ---------------------------------------------------------------------- Date: Sun, 30 May 93 17:08:37 MDT From: Stanton McCandlish Subject: ECON/SPACE: Sell the Universe Whole > The universe owner would eventually break up the universe into smaller > chunks and auction off these chunks, and these new universe-part > owners could similarly auction off smaller chunks. To discourage Umm, well I see a wee little glitch in this: The universe isn't exactly the most finite thing there is. No matter how many subdivisions the original "universe owner" makes and sells off, there is always more, so the original whole-universe buyer has in effect the grandest monopoly of all time. Anyway, I find the whole concept slightly silly. What the hell is wrong with the colonization methods the human race has had for as long as we've been around? I get there first, it's mine. If I can't defend it you can take it away from me. Simple. No need for infinity auctions. -- When marriage is outlawed only outlaws will be inlaws! Stanton McCandlish, SysOp: Noise in the Void DataCenter Library BBS Internet anton@hydra.unm.edu IndraNet: 369:1/1 FidoNet: 1:301/2 Snail: 1811-B Coal Pl. SE, Albuquerque, New Mexico 87108 USA Data phone: +1-505-246-8515 (24hr, 1200-14400 v32bis, N-8-1) Vox phone: +1-505-247-3402 (bps rate varies, depends on if you woke me up...:) ------------------------------ Date: Sun, 30 May 93 19:12:10 EDT From: Andrew S Hall Subject: TECH:Cold feet on cold fusion When the first cold fusion thing came about, I was involved (heavily) in a peripheral way. At that time, most of my friends were physics students and I got sucked into the general frolics. Now, I am really not interested for probably the same reasons as most Extropians. In addition to the reasons outlined by Tim, there is another more compelling reason- the reward is not great enough for the hassle involved. Reading, researching, etc and proving yet another claim to be specious has no rewards in battle like this where partisans will never be convinced. It becomes a quagmire very much like arguing psi research, another topic I avoid. A. Techno-Anarchy.Neophilia.Economic Freedom.Cryptography.Anti-Statism.Personal Liberty.Laissez-Faire.Privacy Protection.Libertarianism.No Taxes.No Bullshit. ********** Liberty BBS 1-614-798-9537 ********** ********** Dedicated to Freedom. Yours. ********** ------------------------------ Date: Sun, 30 May 1993 19:16:14 -0500 From: extr@jido.b30.ingr.com (Craig Presson) Subject: ECON/SPACE: Sell the Universe Whole In <9305300615.AA19226@wookumz.gnu.ai.mit.edu>, Michael Clive Price writes: |> Freeman Craig Presson: |> |> > Find it before you sell it dep't: |> > |> > UPI on 5/26/93 repeats a report from a meeting of the American |> > Geophysical Union in Baltimore to the effect that the Voyager |> > spacecraft, now 4-5 billion miles from the Sun, have begun to detect |> > the predicted LF radio emissions from the heliopause, thus providing |> > our first direct observation of the "boundary" of the solar system. |> |> Do say more.... |> |> What is the heliopause? Is it where the solar wind dies out? And where Yes, the solar wind meets the "cold" gasses outside the solar system and loses any detectable headway, somewhere on the order of 10^10 miles from the sun, according to Ralph McNutt of JHU Applied Physics Lab. My library does not have a recent estimate of the size of the Oort cloud, but dipping back a bit in time I get 50,000 AU = 5x10^12 mi from Whipple in 1974 ("The Nature of Comets" Sci. Am 2/74) and 10^13 miles as an upper bound by A. G. W. Cameron ("The Origin and Development of the Solar System" Sci. Am. 9/75). So it looks like the Oort cloud is O(10^3) larger than the heliosphere. I could definitely use some more recent references on this stuff. Apparently the main mechanism for comets from the Oort cloud to deorbit and come close to the sun is perturbations from the nearer stars. (Talk about your n-body problem!) |> does it lie with respect to the Oort belt? Is this the Voyager craft(s) |> with the infamous 'Sounds of Earth' on? That was Pioneer 10, wasn't it? I think the Voyagers just had a simple plaque. In fact this was a plot device in Star Drek: The Movie way back when: remember "V'ger"? I'll have to check up on this and post on it later; after all, I am a National Space Society and Huntsville Area L5 (HAL5) member now ;-) ^ / ------/---- extropy@jido.b30.ingr.com (Freeman Craig Presson) /AS 5/20/373 P.N.O. /ExI 4/373 P.N.O. ------------------------------ Date: Sun, 30 May 1993 19:25:50 -0500 From: extr@jido.b30.ingr.com (Craig Presson) Subject: HEALTH:Kenneth Seaton & Albumen? In <9305301621.AA21276@wookumz.gnu.ai.mit.edu>, Xiao Zhou writes: |> I just saw Kenneth Seaton on Tony Brown's Journal making some very |> provocative claims about hygiene, health, and albumen. Does anyone |> know the story on this guy? |> It would be difficult for me to summarize his claims. Suffice it to say |> that he presents some interesting opportunities for extropians if he's |> not a crackpot. His claims toward the end began to sound rather extreme, |> and he didn't provide copious citations of where to find these results. |> I didn't see the beginning of the program, so I may have missed something |> important. They did mention some report of the American Heart Assn.(?) I have never heard of the guy, but based on what you say, he sounds so much like a crank that I'd be inclined to discuss making a Metzger bet of say $100 at 1000:1 against his claims. First I'd have to know exactly what the claims were, of course. Yes, that was my $100 to your dime. ^ / ------/---- extropy@jido.b30.ingr.com (Freeman Craig Presson) /AS 5/20/373 P.N.O. /ExI 4/373 P.N.O. ** E' and E-choice spoken here ------------------------------ Date: Sun, 30 May 93 19:00:20 MDT From: Stanton McCandlish Subject: SPACE: Fermi paradox, etc. -- probably dumb question This is likely a "basics" question, but I will ask it anyway, since at the moment I have neither the time nor the funds to go out and buy a research library of space colonization theory books, much less read them all, and I just need a simple quick answer. As I have gleaned from outside reading and this list, the main paradox with the idea that "we are not alone" is that if there were all these civilizations out there, why have they not visited? What'd I'd like to know, is what are the answers to the following questions: 1) Why the hell would they want to come here? 2) Has anyone really given much thought to the SIZE of the universe, and the odds against actually FINDING us? 3) Even after the alpha centaurians or whoever DO find us, who's to say they can afford to come? Who's to say they have the tech to come? Maybe they haven't figured out FTL yet (WE haven't. We have nice theories, but they are hardly reality. Who's to say ANYTHING in the universe has figured it out yet.) Maybe they have short lifespans, maybe they depend upon some resource they can not synthesize or otherwise produce in space? Maybe they can't handle cold, so cryogenic storage for the journey is out of the question? WHO KNOWS? 4) If there are lots of civilizations, some many times as "advanced" as we are who's to say that earth is anything unusual at all? Perhaps we are the space equivalent of an anthill: lots of us, all the same, all boring unless anthills are your specialty. Our SuperAliens might consider us to be nothing but bugs, and hardly worth stopping to look at. 5) Who's to say they aren't already here, or have been here and left. Just because WE don't know it doesn't mean it hasn't happened. All conspiracies- with-the-aliens paranoia aside, this is a serious point to consider. 6) What makes anyone think that "intelligence" as we define it will have any- thing in common with intelligence as THEY define it? Maybe they HAVE been attempting to communicate, and our minds work so different we don't notice. Maybe they are not interested in communicating, and their intellect is enticed by other things. Maybe they are xenophobic, much like 90+% of humanity. Maybe they are sitting patiently in space waiting for the opportune time to invade. Maybe they already DID when we were still not-quite-human hominids, found nothing of interest, left, and never bothered to come back. Maybe they are not in the least interested in spacefaring, and prefer to sit home on Zeta Reticuli II or where ever. Maybe they figure we have some disease that will kill them, much as European diseases wreaked havoc on N. Am. Maybe they are religious nuts, and think earth is Hell, or otherwise an unHoly place to have anything to do with. Etc etc. The main problems I see with the "paradox" are that is relies upon a lot of [to me, very shakey] assumptions, and much of it is based in statistics, which is of limited relevance and usefulness to anything. If this is too basic to be discussed on the list, feel free to reply by email. -- When aliens are outlawed, outlaws will be statistically unlikely! Stanton McCandlish, SysOp: Noise in the Void DataCenter Library BBS Internet anton@hydra.unm.edu IndraNet: 369:1/1 FidoNet: 1:301/2 Snail: 1811-B Coal Pl. SE, Albuquerque, New Mexico 87108 USA Data phone: +1-505-246-8515 (24hr, 1200-14400 v32bis, N-8-1) Vox phone: +1-505-247-3402 (bps rate varies, depends on if you woke me up...:) ------------------------------ Date: Sun, 30 May 93 21:01:59 -0400 From: pcm@cs.brown.edu (Peter C. McCluskey) Subject: New Brain Test: "Einstein einstein.... Penrose!" X-Reposting-Policy: redistribute freely jamesdon@infoserv.com ("James A. Donald") writes: >The scaling variable that I was referring to is the number of degrees of >freedom in the category to be recognized by a category recognition neuron. Finally, a hint of meaning emerges from the sea of noise. However, I still doubt that I understand you, as I don't see any reason to believe that NNs scale up exponentially with what I think you mean by degrees of freedom. It appears to me that once low level processing has achieved rotational, tranlation and scale invariances, that further degrees of freedom are handled with polynomial scaleup under most conditions that humans can handle. >NN theory assumes that high level categories are not generated directly, but >rather are a state of the network as a whole or a large subnetwork. Yet they >are generated directly. Hence my comment that the existence of category >recognition neurons are almost a disproof of NN theory. There are particular The way that some NN theorists have phrased their criticism of the idea of a grandmother cell may have confused some people. NN theory never provided any basis for denying the existence of neurons whose firing correlates reliably with the recognition of a concept, it only said that brains do not depend on any one neuron to recognize a concept (and that claim is independent of the basic theory that ANNs model neurons). >Obviously top down perception is impossible, because of limits of >computational power in conventional AI, and limits on the size of the >"neural net" in NN. Obviously bottom up perception does not work. >Conclusion: Perception is impossible. But obviously it is possible. Top down and bottom up approaches are inadequate. NN researchers who are trying to model a complete visual recognition system use an approach which is primarily bottom up, with feedback from the top levels to lower levels so that the low levels are guided by high level hypotheses. >This is why artificial awareness is not an engineering problem, - not a >question of insufficient power - but a problem of fundamental research - what >the hell is going on here? The fact that there are still some fundamental research problems does not mean the engineering problems are insignificant. ----------------------------------------------------------------------------- Peter McCluskey >> Hofstadter's Law: It always takes longer than you expect, pcm@cs.brown.edu >> even when you take into account Hofstadter's Law. ----------------------------------------------------------------------------- ------------------------------ Date: Sun, 30 May 93 19:38:43 MDT From: Stanton McCandlish Subject: Extropian Currency > * Now, ExI can auction off parts to raise money, which could cut > down on the number of pledge drives. I call "dibs" on: Mars the moons of Jupiter Halley's Comet the rings of Saturn Alpha Centauri and it's vicinity and at least one of the Magellanic clouds. And throw in the nearest dozen black holes and quasars just for kicks. >:) -- When marriage is outlawed only outlaws will be inlaws! Stanton McCandlish, SysOp: Noise in the Void DataCenter Library BBS Internet anton@hydra.unm.edu IndraNet: 369:1/1 FidoNet: 1:301/2 Snail: 1811-B Coal Pl. SE, Albuquerque, New Mexico 87108 USA Data phone: +1-505-246-8515 (24hr, 1200-14400 v32bis, N-8-1) Vox phone: +1-505-247-3402 (bps rate varies, depends on if you woke me up...:) ------------------------------ Date: Sun, 30 May 93 20:51:35 MDT From: Stanton McCandlish Subject: MEDIA: ExI: BBC Radio 4, 0905 2 June 93 > Now why don't they make programmable radio/tape recorders like they do VCRs? They used to. I had a neat little K-martish cheapo AM/FM cassette recorder, ca. 1981, that had a timer, and you could set the tape to start recording whenever. It was totally analog, so you had to set it on the station beforehand. Nowadays, it seems there's just no market for it. Radio is becoming less and less of a force in the world it seems...BUT some highend stereo systems can do this. My mother's exhusband has a BIG Kenwood system that will do this sort of thing. Just seems the cheapo models won't do it anymore. BUT! Maybe there IS a market for this, especially if advertised correctly? Perhaps some of the more entrepreneurial among you could find a niche making programmable boomboxes? -- When marriage is outlawed only outlaws will be inlaws! Stanton McCandlish, SysOp: Noise in the Void DataCenter Library BBS Internet anton@hydra.unm.edu IndraNet: 369:1/1 FidoNet: 1:301/2 Snail: 1811-B Coal Pl. SE, Albuquerque, New Mexico 87108 USA Data phone: +1-505-246-8515 (24hr, 1200-14400 v32bis, N-8-1) Vox phone: +1-505-247-3402 (bps rate varies, depends on if you woke me up...:) ------------------------------ Date: Sun, 30 May 93 22:24:46 CDT From: derek@cs.wisc.edu (Derek Zahn) Subject: fetishism? So... today I upgraded. As anybody who read all the way through the latest Exponent knows, I'd been running a 486-SX20 (yes, they still exist, though I don't know if they can still be bought). Today I installed the New Thing... the local bus DX2-66... the Monster. Now, I'm sure that many people have increased their personal computing resources without boring the list with that experience -- but to me this is an Event. It includes some interesting bits: * Where did I buy it??? First, the brand name is "Pionex". I had never heard of that name; I studied the tech specs and came out with the thumbs-up, but still it's a nothing to me. I'd thought that I'd buy out of Computer Shopper, but then decided that it's only fair to check out the local talent. The first thing I found was a mind-boggling deal from (yes) Sears -- 486DX-33 local bus 4MB 70ns RAM, 245 MB disk [anybody but me noticed the SUDDEN plunge in disk prices last few mos?] normal 1024x768 NI 14" monitor -- all for $1499. 'Hey' I said. 'Shit'. That's a damn good price. But Mr. Sears had no 66, which is what I wanted. So I rethought the M.O. default scenario and went to malls (love malls, gotta love malls). Found the Pionex thing... dx2-66, 4mb (plus another 4mb scavanged immediately from the Old Hag), 200+mumble HD, + 1024x768 display NI. Price: $1799. Not At All Bad. For the record, got it as Sam's Club (poor dead Mr. Walton...). * To recap: Check out the local superstores before assuming that mail order is the way to go. I could have got a similar setup for about $50 less mail order, but the benefit of carry-out (and carry- back if it was non-functional) was important to me. * I would like to re-OOMPH the usual points about increasing technologies re: computers... sure; pentium, hexagon, ... -- but the last machine I personally paid for (as opposed to having consulting firms buy me) was an Amiga 1000 circa 1985. About the same price, no hard drive, ... ... ... yowza! I (along with Stanton) am in the process of starting an independent software company, in which I intend to put to good use the abstract AI stuff I've learned. If there are any out there besides just Stanton and I, I'd be interested in an Extropian businessmans' forum. In case anybody's curious (NOT DISTRIBUTABLE BEYOND EXTROPIANS): my business area: After LONG and careful thought, I've decided that my first area of product development would be Speech Recognition. I may abandon my current efforts toward a PhD in order to give it proper attention. I believe that it will be a MANY billion dollar area in the next decade and that I can have a piece of that pie, because my training and inclination are all in this direction (maybe I can't have _much_ of the multi-billion, but then I don't need much...)... I'll keep y'all informed as things progress... I've rarely felt so _alive_... derek ------------------------------ Date: Sun, 30 May 93 22:01:13 MDT From: Stanton McCandlish Subject: META: Split the list I personally like the idea of separate lists, and in fact proposed this myself about a month or 2 or 3 ago, I forget. More interesting to me is Derek's idea of a Extropian businessfolks' list. It would be a Good Thing I think to have several specialty lists. Then people not interested in Randian philosophistry don't have to subscribe to exi-rand@gnu.ai.mit.edu, and can stick to exi-ftl@gnu.ai.mit.edu. You get the idea. -- When marriage is outlawed only outlaws will be inlaws! Stanton McCandlish, SysOp: Noise in the Void DataCenter Library BBS Internet anton@hydra.unm.edu IndraNet: 369:1/1 FidoNet: 1:301/2 Snail: 1811-B Coal Pl. SE, Albuquerque, New Mexico 87108 USA Data phone: +1-505-246-8515 (24hr, 1200-14400 v32bis, N-8-1) Vox phone: +1-505-247-3402 (bps rate varies, depends on if you woke me up...:) ------------------------------ Date: Sunday, 30 May 1993 20:11:37 PST8 From: "James A. Donald" Subject: Chaitin, Holland, & AI Twenty years ago this was a feeble excuse. Now it is an old and feeble excuse. In <9305301728.AA21591@wookumz.gnu.ai.mit.edu>, szabo@techbook.com (Nick Szabo) wrote: > * The fact that traditional algorithm-complexity theory shows a computation > to be NP-complete for the exact solution doens't mean that a very good > approximate solution cannot be found in a vastly shorter time, eg via > genetic algorithms which eliminate O(k^n/n) possibilities per step. > * The vast bulk of practical human intelligence involves finding > approximate solutions. Approximate solutions to the best move in chess are good moves. Approximate solutions to interpreting sensory data as object data are wrong interpretations. Perception involves finding a description of sensory data in terms of objects and entities. You do not want a *good* interpretation - one with *few* absurdities and impossibilities. You want the *right* interpretation - the one with *no* absurdities and impossibilities. In other words: The areas where it is hard for AI to emulate the capabilities of animals, are precisely those areas where the problem really is hard. The areas where we have had least success are precisely those areas where approximate solutions are nonsense solutions. In most areas where approximate solutions are valid, such as chess, computers do very well. This rationalization about approximate solutions always brings up the joke about the drunk looking for his keys: A drunk is carefully examining the ground under a streetlight. A passerby asks him what he is doing. The drunk replies: "I am looking for my keys." The passerby looks around. No keys in sight. So he asks: "Do you remember where you lost your keys?" Drunk: "Sure I do. I went into the bushes over there to take a leak, and while I was fiddling with my pants my keys fell into the long grass." Passerby: "Then why are you looking for your keys here instead of over there?" Drunk. "No hope of finding them over there, it's completely dark." Ninety five percent of the work done in AI. and ninety nine percent of the work done in neural networks consists of looking for keys under the streetlight. The key to artificial awareness is not under the streetlight. Unfortunately it is in the dark under the long grass. Perception is non polynomial by all known methods, but is not NP complete. But no one has been able to find a useful difference between perception and problems that *are* NP complete. If we could find such a useful difference, then possibly we could make some progress. Without such a discovery, algorithmic methods, including NN, will necessarily go nowhere. > Clearly functional AI has succeeded in many areas, especially if > we include the broad sweep of computer science under this rubric > (as I think we should; all of computer science is concerned with > implementing and extending aspects of human intelligence on computers). Twenty years ago it became apparent that perception was the major roadblock. In those days the problem was called the combinatorial explosion, not NP, but though the terminology has changed no progress has been made on this matter. The key roadblock is completely unchanged. No progress whatever, other than a better understanding of how the road is blocked, and more powerful tools to find out how animals do the impossible. Computers are doing just great at doing those things that humans can do but other animals cannot. They can often do such things a lot better than humans. It is in precisely those matters where simple animals can perform feats similar to humans, that computers fail dismally. This in itself suggests that the size of the brain is not terribly important in those matters, hence computational power is not important. > Modeling AI has largely failed to model human intelligence in terms > of traditional Platonic computer science or AI research; better > results are being had with more biologically faithful > models like neural nets. I will believe that when I see a neural net that can handle the origami microworld. So far conventional AI scores 0.01 percent out of 50 percent required for a pass, neural nets 0.001 percent. Conventional AI may fail dismally when compared to a spider, but it still leaves NN in the dust. --------------------------------------------------------------------- | We have the right to defend ourselves and our James A. Donald | property, because of the kind of animals that we | are. True law derives from this right, not from jamesdon@infoserv.com | the arbitrary power of the omnipotent state. ------------------------------ Date: Sunday, 30 May 1993 20:20:12 PST8 From: "James A. Donald" Subject: New Brain Test In <9305302041.AA23022@wookumz.gnu.ai.mit.edu>, price@price.demon.co.uk (Michael Clive Price) wrote: > > James A. Donald says: > > > As I said [..] the scaling variable is .. > > The original claim: > > >>> Since the amount of processing power required goes up > >>> nonpolynomially fast with the scaling variable, (in the > >>> case of "neural nets" the size of the neural net") My statement was ambiguous. As I made clear in my subsequent clarification, what I meant was that in the case of neural nets the processing power is the size of the neural net. What I meant to say, and subsequently did say was: Since the amount of processing power required goes up nonpolynomially fast with the scaling variable, (in the case of "neural nets" the size of the "neural net" goes up non polynomially fast with the scaling variable.) .... The scaling variable is the number of degrees of freedom in those category detection neurons whose categories possess the largest number of internal degrees of freedom. > When a NN is too large it switches from categorical learning into rote > learning. The required size is before rote learning takes over, so that > it's forced to meaningfully compress its microworld model. What is > remarkable is the complexity of the microworld than can be processed by > simple NNs. The microworlds that I have seen are pretty small compared with the microworlds that conventional AI did fifteen years ago. Have neural nets done the origami microworld? How did they go on the negative chair illusion - do they make the error that they should make? > > The problem is [...] a great many neurons whose output represents a > > wide variety of high level categories, for example "face of adult > > male", or the face of a particular individual. NN theory assumes that > > high level categories are not generated directly, [..] Where are the > > intermediate level abstractions? > > Absence of evidence is not evidence of absence. True. But we would expect intermediate level categories to be vastly more numerous than readily intelligible categories. It is curiously easy to find readily intelligible categories, which suggests that there are few if any intermediate stages between a low level readily intelligible features and high level readily intelligible categories. --------------------------------------------------------------------- | We have the right to defend ourselves and our James A. Donald | property, because of the kind of animals that we | are. True law derives from this right, not from jamesdon@infoserv.com | the arbitrary power of the omnipotent state. ------------------------------ Date: Sunday, 30 May 1993 20:37:10 PST8 From: "James A. Donald" Subject: New Brain Test: "Einstein einstein.... Penrose!" In <9305310102.AA25893@wookumz.gnu.ai.mit.edu>, pcm@cs.brown.edu (Peter C. McCluskey) wrote: > Top down and bottom up approaches are inadequate. NN researchers who > are trying to model a complete visual recognition system use an approach > which is primarily bottom up, with feedback from the top levels to > lower levels so that the low levels are guided by high level hypotheses. Conventional AI tried that a long time ago. In practice it reduces to conventional top down, unless you impose arbitrary timing limits in which case it reduces to conventional bottom up. Furthermore that does not seem to be the way that real animals solve perception problems. Bottom up with feedback from high level hypotheses involves endless cyclic chugging around. It is clear that animals solve many perception problems so rapidly that the flow of information must be predominantly unidirectional in most cases. Non human animals do not need to think things through. Only computers, humans, and perhaps some non human great apes do that. Animals perceive as swiftly as only pure bottom up methods can, but they are as flexible as only almost pure top down methods can be. > The fact that there are still some fundamental research problems does not > mean the engineering problems are insignificant. I agree with that point completely. I am not saying that work on NN is worthless, I am saying let us hold down the ambitious and ambiguous claims that gave AI such a well deserved bad reputation a decade or two ago. In particular the claim that NN works the way real brains work remains to be proved. They work the way real brains work in those cases, such as simple conditioned reflexes, where real brains do not work particularly well. What people wish to see is them working the way real brains work in those cases where real brains *do* work well. Let us also consider that the keys are not under the streetlight, and avoid once again going over ground that proved barren a long time ago. > ----------------------------------------------------------------------------- > Peter McCluskey >> Hofstadter's Law: It always takes longer than you expect, > pcm@cs.brown.edu >> even when you take into account Hofstadter's Law. > ----------------------------------------------------------------------------- --------------------------------------------------------------------- | We have the right to defend ourselves and our James A. Donald | property, because of the kind of animals that we | are. True law derives from this right, not from jamesdon@infoserv.com | the arbitrary power of the omnipotent state. ------------------------------ Date: Mon, 31 May 1993 00:27:53 -0600 From: twb3@midway.uchicago.edu (Tom Morrow) Subject: PGP: A paranoid scenario Although crypto fans often tend toward paranoia, I have not yet seen anybody discuss the following scenario. Statist spooks don't like PGP, but find they cannot realistically spare the computational power required to crack it. The can, however, spare the computational power required to search internet for messages containing strings like "-----BEGIN PGP PUBLIC KEY BLOCK-----" and to store these messages away for later processing. The spooks figure that given the exponential growth in the amount of computing power they can get for each dollar, they will be able to decode these stored PGP messages soon enough. Soon enough, that is, to beat the statute of limitations on whatever "crimes" they uncover. It would be interesting to see some guesstimates on how much memory the spooks would have to have to store all the PGP-encoded messages currently sent on the net, and on how quickly they will be able to decipher them. True paranoids will really only care about the second figure, however, since if the spooks are storing anyone's PGP-encoded messages paranoid folks will figure that it's *their* messages that the spooks are saving. T.O. Morrow -- twb3@midway.uchicago.edu Vice President: ExI -- The Extropy Institute Law & Politics Editor: EXTROPY -- Journal of Transhumanist Thought ------------------------------ Date: Sun, 30 May 93 22:58:46 -0700 From: davisd@pierce.ee.washington.edu Subject: Chaitin, Holland, & AI > From: "James A. Donald" > > Perception involves finding a description of sensory data in terms of objects > and entities. You do not want a *good* interpretation - one with *few* > absurdities and impossibilities. You want the *right* interpretation - the > one with *no* absurdities and impossibilities. Phooey. You can have lots of absurdities, as long as they don't prevent you from completing whatever task you have. Can you say perceptual illusions? You also might wish to consider that there may be more than one interpretatiowithout absurdidtes and ambiguities. Buy Buy -- Dan Davis ------------------------------ Date: Mon, 31 May 93 08:25:40 BST From: Richard Kennaway Subject: Extropian Death Rituals Slightly tangential to Habs' posting, I heard this morning an item on a radio news programme about Japanese death rituals -- for computer chips. Apparently, at the end of each year, some electronics companies hold a Buddhist ritual to make peace with the spirits of all the chips they sacrificed during the year: defective chips, test samples, preliminary versions, and so on, all the stuff that never gets out the front door. If these chips have spirits, they might hold a grudge, so the service is held to thank them and bid them farewell. Everyone *sounded* like they were keeping a straight face. -- ____ Richard Kennaway __\_ / School of Information Systems Internet: jrk@sys.uea.ac.uk \ X/ University of East Anglia uucp: ...mcsun!ukc!uea-sys!jrk \/ Norwich NR4 7TJ, U.K. ------------------------------ Date: Mon, 31 May 1993 04:06:35 -0500 (EDT) From: rjc@gnu.ai.mit.edu Subject: fetishism? Derek Zahn writes: > > > So... today I upgraded. As anybody who read all the way through the Yeah!! "I feel the need, the need - for SPEED!" > * I would like to re-OOMPH the usual points about increasing > technologies re: computers... sure; pentium, hexagon, ... -- but > the last machine I personally paid for (as opposed to having > consulting firms buy me) was an Amiga 1000 circa 1985. About the > same price, no hard drive, ... ... ... yowza! Poor ole A1000, it was a great hacker's machine for the time. You can get an Amiga 3000 for $900 now. BTW, The designers of the A1000 are at it again (Jay Miner, Dave Needle, Leo Schwab, Rj Mical), they designed a new machine called the 3DO. You can read about it in most games/video rags (RISC processor, dual speed CD rom, awesome graphics/audio custom chips, lots of industry backing. Supposed to be a CD-I killer.) > I (along with Stanton) am in the process of starting an independent > software company, in which I intend to put to good use the abstract > AI stuff I've learned. If there are any out there besides just Stanton > and I, I'd be interested in an Extropian businessmans' forum. > (NOT DISTRIBUTABLE BEYOND EXTROPIANS): > > my business area: > > After LONG and careful thought, I've decided that my first area of > product development would be Speech Recognition. I may abandon > my current efforts toward a PhD in order to give it proper attention. > I believe that it will be a MANY billion dollar area in the next > decade and that I can have a piece of that pie, because my training > and inclination are all in this direction (maybe I can't have _much_ > of the multi-billion, but then I don't need much...)... I don't know if there's much of a market there for small time players, unless you can beat the stuff that Apple and AT&T is working on. I heard that Apple intends to market a system for incorporating voice recognition into consumer devices (TV, VCR, Microwave, etc). Apple's system is speaker independent. (these are their claims, although I've never seen them in action) All those voice recognition VCR controllers will be killed when speaker-independent chips/systems hit the market. Usually, you have to "teach" the controller to understand your voice, and then it will only recognize yours. Not very user friendly. Since Apple's system will be moderately cheap and good at recognizing speed there may not be a market for voice recognition software on PCs/Macs or embeded controllers. That is, if this voice recognition is provided as a standard operating system object, there won't be much use for third party software unless it is significantly better. I bet Microsoft will also announce or buy voice recognition software for Windows after Apple releases their product. I know that voice recognition still has a long way to go, but a a speaker independent system that's moderately good will sell well. Small improvements on that won't penetrate that much. Since development has been going on for a long time, you may have chosen to enter the market too late. (isn't Apple supposed to come out with its technology this summer? Their new machines are supposed to have DSPs in them and I bet that's not just for audio effects. AT&T is also working on something. I tested a demo of it that they had on an 800 number about a year ago and it worked ok. [recognized all the numbers, one, two, three, etc. yes/no, and certain commands]) -- Ray Cromwell | Engineering is the implementation of science; -- -- EE/Math Student | politics is the implementation of faith. -- -- rjc@gnu.ai.mit.edu | - Zetetic Commentaries -- ------------------------------ Date: Mon, 31 May 1993 04:14:19 -0500 (EDT) From: rjc@gnu.ai.mit.edu Subject: New Brain Test: "Einstein einstein.... Penrose!" Peter C. McCluskey writes: > >This is why artificial awareness is not an engineering problem, - not a > >question of insufficient power - but a problem of fundamental research - what > >the hell is going on here? > > The fact that there are still some fundamental research problems does not > mean the engineering problems are insignificant. Also, with faster hardware you can do faster research. Think of how computing used to work before PCs -- batch-processing. You'd have to wait hours just to find out if your program compiled correctly, and the output might have taken even longer. Now the edit-compile-debug cycle is down to seconds. (my professors just love talking about painstakingly keypunching a fortran program and waiting the next day to see if their physics/engineering solutions worked! They also babble about how faster computers have caused longer work hours. They say that back in the 60s, you could batch-process your program then go on a coffee break/luncheon. Nowadays, you don't have an excuse for not being chained to your keyboard) With high speed hardware you can train/run NNs faster. Therefore the hypothesis->result->refinement cycle works faster. Conclusion: faster hardware will greatly help the research effort. > ----------------------------------------------------------------------------- > Peter McCluskey >> Hofstadter's Law: It always takes longer than you expect, > pcm@cs.brown.edu >> even when you take into account Hofstadter's Law. > ----------------------------------------------------------------------------- > -- Ray Cromwell | Engineering is the implementation of science; -- -- EE/Math Student | politics is the implementation of faith. -- -- rjc@gnu.ai.mit.edu | - Zetetic Commentaries -- ------------------------------ Date: Mon, 31 May 93 02:11:38 -0700 From: davisd@pierce.ee.washington.edu Subject: Speech recognition Derek Zahn writes: > After LONG and careful thought, I've decided that my first area of > product development would be Speech Recognition. I may abandon > my current efforts toward a PhD in order to give it proper attention. > I believe that it will be a MANY billion dollar area in the next > decade and that I can have a piece of that pie, because my training > and inclination are all in this direction (maybe I can't have _much_ > of the multi-billion, but then I don't need much...)... I don't know if there's much of a market there for small time players, unless you can beat the stuff that Apple and AT&T is working on. I agree. I wish you luck, but a lot of folks are spending oodles of time and money to do this, and my understanding of the state of the art is that a lot of it is painstakingly handcrafted. Probably too big a project for a small startup to be successfull in. If, however, you have some particular new approach you plan on bringing to the area, you might be able to get research funding for it. What are the general outlines of your business plan? From: rjc@gnu.ai.mit.edu I bet Microsoft will also announce or buy voice recognition software for Windows after Apple releases their product. Microsoft is making a big push toward voice recognition. They have hired the architect of the Sphinx system, no doubt at considerable cost, to help their efforts. (Speech recognition is not my area of expertise, but a couple of guys who work under my advisor are working in the area. They talk, I listen, I pass it on. My advisor was of the opinion a year or two ago that it was much too late for a small outfit to get in on the action, at least as far as developing a real product goes.) Buy Buy -- Dan Davis ------------------------------ Date: Mon, 31 May 93 04:13:09 MDT From: sneal@muskwa.ucs.ualberta.ca (Sneal) Subject: Fetishism/Speech Recognition/Extropian Business Bureau Derek Zahn writes: > The first thing I found was a mind-boggling >deal from (yes) Sears -- 486DX-33 local bus 4MB 70ns RAM, 245 MB disk >[anybody but me noticed the SUDDEN plunge in disk prices last few mos?] >normal 1024x768 NI 14" monitor -- all for $1499. 'Hey' I said. 'Shit'. >That's a damn good price. Almost exactly two years ago, I paid US$2600 for the same system with a 120 Mb drive... after spending countless hours dickering with Computer Shopper advertisers and shrewd Korean computer merchants. mutter mutter lousy stinkin sears mutter gripe > Found the Pionex thing... dx2-66, 4mb (plus another 4mb scavanged >immediately from the Old Hag), 200+mumble HD, + 1024x768 display NI. >Price: $1799. Not At All Bad. So.... are there specific times when you're not at home? Do you have any other consumer electronics in the household? For my part, I gave in to the siren song of multimedia this weekend and sprung for a SoundBlaster 16 ASP and a Mitsumi CD-ROM drive (no doubt the prices of same will be cut in half next week). Once I've figured out what I want for video input, you all can look forward to the much-talked-about "Extropian Beach Party" movie in MPEG format. Keep your surfboard wax dry, everyone. >I (along with Stanton) am in the process of starting an independent >software company, in which I intend to put to good use the abstract >AI stuff I've learned. If there are any out there besides just Stanton >and I, I'd be interested in an Extropian businessmans' forum. Count me in. After several years of developing telecomm software (and gradually losing ground to the copycats as they get ever faster at copying designs) I'm itching to get into something on the leading edge or thereabouts. After learning that exotic matter has nothing to do with exotic dancers, I tossed out my traversable wormhole business plan, and now that Robert McElwaine has lost his net access, I *need* a new source of brilliant but unconventional ideas. >After LONG and careful thought, I've decided that my first area of >product development would be Speech Recognition. Hey, does anyone know what happened with Raymond Kurzweil's efforts in this area? I read something long ago (in Byte, I think) about a medical vox recognition system RK was marketing for surgeons and pathologists. Anything new happening? -- Steve -------------------------------------------------------------------- "Can you imagine - a world without lawyers?" - Lionel Hutts, CEO, 'I Can't Believe It's a Law Firm' ------------------------------ End of Extropians Digest V93 Issue #0291 ****************************************