From extropians-request@gnu.ai.mit.edu Fri Jun 4 14:26:16 1993 Return-Path: Received: from usc.edu by chaph.usc.edu (4.1/SMI-4.1+ucs-3.0) id AA02993; Fri, 4 Jun 93 14:25:43 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 AA23736; Fri, 4 Jun 93 14:24:11 PDT Errors-To: Extropians-Request@gnu.ai.mit.edu Received: by wookumz.gnu.ai.mit.edu (5.65/4.0) id ; Fri, 4 Jun 93 17:19:37 -0400 Message-Id: <9306042119.AA18450@wookumz.gnu.ai.mit.edu> To: ExI-Daily@gnu.ai.mit.edu Date: Fri, 4 Jun 93 17:19:16 -0400 X-Original-Message-Id: <9306042119.AA18436@wookumz.gnu.ai.mit.edu> X-Original-To: Extropians@gnu.ai.mit.edu From: Extropians-Request@gnu.ai.mit.edu Subject: Extropians Digest V93 #0299 X-Extropian-Date: Remailed on June 4, 373 P.N.O. [21:19:36 UTC] Reply-To: Extropians@gnu.ai.mit.edu Errors-To: Extropians-Request@gnu.ai.mit.edu Status: OR Extropians Digest Fri, 4 Jun 93 Volume 93 : Issue 0299 Today's Topics: AI: Cyc & the SAT wager [1 msgs] EXTROPRENEUR: Request for Comments [1 msgs] FUQ? Some CyC info... [1 msgs] FWD: To destroy an anarchy (fwd) [1 msgs] PHYS/COMP: Quantum Computer Feasible Now?! [1 msgs] PREFIXES: The 'README' meme [1 msgs] Power Crazed FDA [2 msgs] SPACE: Fermi paradox, etc. [3 msgs] TECH: DAK = trash [1 msgs] TECH: Source for shortwave w/ cassette recorder [1 msgs] TECH: shortwave [2 msgs] a new seminar series [1 msgs] help Extropians [1 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: 50384 bytes. ---------------------------------------------------------------------- Date: Fri, 4 Jun 1993 12:11 EDT From: Ronald Hale-Evans Subject: TECH: Source for shortwave w/ cassette recorder >Catalog item #5769 $79.90 + $6 P&H 800-DAK (325)-0800 Anytime > >Dual stereo cassettes (1 records, both play) 530-1630kHz (AM) >Programmable timer 3200-7300kHz (Shortwave1) >20 presets for tuner 9500-21750khz (Shortwave2) >etc. 87.5-107.9MHz (FM/Stereo) WARNING!!! This equipment is a piece o' shit. DO NOT BUY IT. I made that mistake, and within about forty hours of usage, the following things went wrong with it: * The LCD display became impossible to read (this effect was temporary, but a real nuisance). * Both cassette drives failed. * The unit will no longer run from battery power. * The unit will no longer take power from the battery that powers the memory, so you can't preset stations and you can't keep the clock set. You must set both of these every time you turn on the unit. As it turns out, I am not alone in observing that this unit is crap. When I got it, I signed up on a shortwave mailing list, and many people told me they had had the same bad experience with DAK shortwave equipment. The DAK warranty is a ripoff in my opinion, so you won't get any satisfaction from that. * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * "The Sea refuses no river; remember that when a beggar buys a round." --Pete Townshend * * * * * Ron Hale-Evans, evans@binah.cc.brandeis.edu PGP 2 public key: finger evans@binah.cc.brandeis.edu ------------------------------ Date: Fri, 4 Jun 1993 08:55:34 -0700 (PDT) From: szabo@techbook.com (Nick Szabo) Subject: AI: Cyc & the SAT wager In an earlier AI thread, I argued that AI has not "failed", and offered to wager that a reasonable-sized project (one professor, one postdoc and a few grad students over four years) combining Cyc, genetic programming, and a few standard software packages (eg Mathematica) could produce a program that would score better than most humans on standard intelligence tests used by universities, MENSA, etc.: the SAT and GRE. I also declared that the AI community is scared to death that university-style intelligence will be the first to be automated, since the traditional "robots throwing blue collar workers out of work" scenario is more likely to be applied to scientists and engineers. Thus, most AI researchers will subconsciously try everything under the sun to avoid throwing this bombshell. (This is also why one of the wager conditions is that I run the project). Here follows a description of Cyc. Note that the "final test" for Cyc involves no such standardized intelligence test, even though it would be the most natural kind of "final test" for an "artificial intelligence" project. Note also the emphasis on analogies, which might be the most difficult part of the SAT and GRE tests for a computer program. I do have some quarrels with the Cyc project. I object to the notion that Cyc somehow defines "consensus reality". I hope to see a wide, diverse variety of competitors for Cyc in the coming years. Using genetic programming, I think its early knowledge-aquisition phase could be greatly automated. Nick Szabo szabo@techbook.com Forwarded message: Date: Thu, 3 Jun 1993 20:05:17 -0600 (MDT) From: pfly@nyx.cs.du.edu (juxlus.) To: Aleph@pyramid.com (Aleph) Subject: FUQ? Some CyC info... Message-Id: <9306040205.AA16024@nyx.cs.du.edu> _Building_Large_Knowledge_Based_Systems_: Representation and Inference in the Cyc Project, Douglas B. Lenat and R.V. Guha, Addison-Wesley, Reading, Mass: 1990. CyC is a project ebing run by Microelectronics and Computer Technology Corporation (MCC) since 1984. It is a ten year project, scheduled to be "finished" by 1994. pg. 3-4 "Why aren't human beings brittle in the way computer programs are? How do we cope with novelty? Largely by finding some related case and propagating the differences to this new one. That is, we do one of the following: o Match some similar situation and adjust it slightly (= remember) o Match some apparently far-flung situation (= analogize) o Fall back on general knowledge (= use common sense) o Try to learn more about this new situation (= recur) That fourth case is perhaps a special case of the third. In any event, it is a *recursion*, in which our "problem" changes from "X" to "learn more about X", and any of the above four methods can now be tried on that new problem... ...The first three cases above (and hence, ultimately, the fourth) depend on having a large base of both general and specific knowledge to consult... ...So the matress in the road to AI is lack of knowledge, and the antimattress is knowledge. But how much does a program need to know to begin with? The annoying, inelegant, but apparently true answer is: a non-trivial fraction of *consensus reality*---the millions of things that we all know and that we assume everyone else knows... ...the Cyc group at MCC is attempting to build a single intelligent agent whose knowledge base contains these tens of millions of entries. We believe such a system will be a useful, perhaps necessary, platform on which to undertake the next generation of work in expert systems, natural language understanding, and machine learning." pg. 10-11 "The previous subsection argued that analogy is frequently useful and discussed speciafically *how* it might be useful. But *why* is it frequently useful? The answer to that lies inb the nature of the world we happen to inhabit and the state of our understanding (and perhaps our capability to understand). Three aspects of our world, and ourselves, make analogy frequently useful to us as human beings: 1. The moderate amount of novelty with which we're confronted 2. The moderate number of distinct causes in the world 3. The medicore entology and "knowledge metric" we have NOVELTY. If the world were very volatile and chaotic and wildly unpredictable, analogizing would avail us little; it would almost always fail. If it were totally staid and static and unchanging, we'd have little need for analogy; memory would suffice. CAUSALITY. Analogies that have no common root cause are superficial and weak and, more often than not, no good for mroe than literary divertissement. If there were a zillion disparate causes in the world, analogy would usually be superficial and powerless in this fashion. On the other hand, if there were no variety of causal mechanisms in the world (for example, only one emotion, only one method of physical propulsion, only one kind of chemical reaction, etc), there wouldn't be much power in classifying which of thsoe causes were behind something. (In fact, tghere wouldn't be much point in having those terms either.) KNOWLEDGE METRIC. If we had a terribly wrong view of the world, analogy would lead us even further into error (for example, thinking that the cosmic objects and meteorlogical phenomena are sentient, one might try to bribe them with offerings). If we had a terrific grasp of the world, we'd always *know* precisely what knowledge was relevant to our present dilema, and exactly how it should be applied. We wouldn't need analogical leaps into the possibly relevant." pg. 15 "The careful reader may have noted a possible hole in our discussion of the brittleness of current expert systems: First, we criticized them for merely containing opaque tokens and pushing them around. Yte our example of "having more general, flexible knowledge" was nothing more than having more (and more general) tokens and pushing *them* around! Yes, all we're doing is pushing tokens around, but that's all that cognition is. What makes our tokens "better" is that they aren't tied to solving some particular problem. Naturally, all programs are built on some primitives (predicates, frames, slots, rules, functions, scripts). But if you choose task-specific primitives, you'll win in the short run... but lose in the long run (you'll find yourself painted into a corner when you try to scale the program up)." pg. 17 "In knowledge representation...narrow domains are misleading. And small KBs likely won't scale up easily into huge KBs. This, then, is the representation trap, the trap that has snared (or even *characterized*) expert systems to date: choosing a set of long, complex primitives (prediacte names) that have a lot of knowledge compiled within them, and writing rules that are also tailored to the program's domain (omitting premises that needn't be worried about in that particular task). The bait in the trap is the fact that it works---at least within the narrow domain for which that particular program was designed. The catch is that the resultant system is isolated and brittle." pg. 20-21 "...Instead of talking more about this, and making forays here and ther as needed to gather a few examples, we set out to actually do it. Well, almost. We figured that if we could get pretty far on setting up the top layers, everyone would pitch in and help us relate the rest of human knowledge to them. Is it possible? Who knows? But let's get started and see! That was our attitude in 1984; we gave Cyc a 10-20 percent chance of succeeding. Now it looks like our original guess about the size of the task was about right, and we'd give us a full 50-60 percent chance of succeeding. The first task involves on the order of ten million entries (As Marvin Minksy observed, that's about the same order of magnitude as the number of things a human being acquires---burns into long term memory---during the ages 0 to 8, assuming one new entry every ten second's of one's waking life). The second task is unbounded, but probably another ten to fifty million entries would suffice for general intelligence (for example, the intelligence required for acquiring knowledge in school and in extra-curricular conversations). Quite a bit more may be needed for quantatively super-human intelligence. The most easily foreseen mode of failure was---and is---that the knowledge enterers might diverge, either by stepping on one another's toes (misusing each other's terms) or by passing one another in the night (re-entering already existing concepts, giving those units slightly different names). If the latter duplication is ever noticed, then it may be fairly easy to fix (either by related the units to one another, or, in extreme cases, just merging them), so the failure mode would lie in never realizing that this duplication occurred in the KB. One interesting tool that helps in identifying such duplication is to have Cyc actively search for new analogies. Some of them are genuine, interesting analogies; some of them are mappings between non-analogous concepts A and B, which signifies that we haven't yet told Cyc enough about A and B to differentiate them properly; and a third class of apparent analogies are between concepts that are really just different formulations of the same knowledge---that is, passings in the night. Since 1984, we've been building and organizing and reorganzing our growing consensus reality KB in Cyc. We now have about a million entries in it. and we expect it to increase by a factor of 4 by mid-1990. Thanks to an array of explicit and implicit methods for stating and enforcing semantics, they appear to be converging, not diverging." pg. 23 "So we must build a good global ontology of human knowledge (that is, one that spans current human consensus reality) if we are to avoid the representation trap. Choosing a set of represenation primitives (predicates, objects, functions) has been called *ontological engineering*---that is, defining the categories and relationships of the domain (This is empirical, experimental engineering, as constrasted with *ontological theorizing*, which philosophers have done for millennia). More than just having a good ontology, however, we must also build up a large knowledge base organized according to that ontology, a KB of millions of (frame-sized) pieces of consensus reality knowledge. How many millions? We hope and expect it's about 5 million frames, each with several dozen "fact-seized" slot entries; but we'll find out! This project is mankind's first foray into large-scale ontological engineering." pg. 26-27 "... We are building the needed KB manually, one piece at a time, at least up to the crossover point where natural language understanding begins to be a more effective way of further enlarging it. This task tlooked---and still looks---just barely possible. We estimated that it would take about two person centuries to build up that KB, assuming that we don't get stuck too badly on represenation thorns along the way. In real time, our schedule was---and still is---to complete the project (reach that crossover point) in the ten-year peroid that ends in late 1994." pg. 28 "Cyc comprises three "pieces": 1. The knowldge base itself (Cyc KB) 2. The environment: the interface editing/browsing tools (UE and MUE), the multi-user knowledge server, the binary KB dumper, and so on. 3. The represenation language (CycL)" pg. 35 "Superficially, CycL is a frame-based language; that is, it's based around triplets like "the *capital* of *Texas* is *Austin*". All the assertions about Texas are gathered into one data structure called the Texas *frame* or the Texas *unit*: Texas capital: (Austin) residents: (Doug Guha Mary) stateOf: (UnitedStatesOfAmerica) The unit representing Texas is depicted here as having three *slots*, each of which has a corresponding *value*. The value of a slot of a unit is always a list of individual *entries*..." pg. 37 "Four basic kinds of frames exist in the system: "normal" ones, SlotUnits, SeeUnits, and SlotEntryDetails. pg. 45-46 "The language [CycL] can be conceptualized as being formed by four important modules (of vastly varying size): o *The inferencing module* that is responsible for interpreting the rules available in Cyc (both monotonic and default "rules"). In other words, given that a "rule" *might possibly* be relevent, this module sees if it is indeed relevant in the current state of the KB. If it is relevant, this module "runs" it to produce some conclusion, some new assertations. o *The stripped-down truth maintenance system (TMS) conceptual module*. Suppose a rule ran, and concluded u.s.v1. Since then, we've learned some more things, the KB has changed, and now it turns out that that rule would no longer have run. If u.s.v1 is till around, it has become 'stale'; all its support has gone away. This TMS module ensures that u.s.v1 is retracted if there is no other support for it. This module is referred to as a *conceptual* module because as we shall see later, this module has been partially unified with the inference module. However, for purposes of understanding the design of the language, it is convienent to regard this as a seperate module. o *The contention resolution module* that compares the conclusions of different rules and, in case of conflict, attempts to resovle the conflict. o *The contradiction detection and resolution module* that detects situations when the conclusions of monotonic (absolutely certain) rules have been overridden by those of other monotonic rules. The only way to resolve this kind of contradiction, short of retracting or weakening one of the rules, is to change the KB so that the antecedent of one of the contradictory rules is not satisfied. This module detects such contradictions, and, when it is possible and reasonably safe to do so, automatically modifies the KB to preserve consistency." pg. 135 "A *task* is a script that can be performed by the system. Each kind of task must be an instance not just of Script (= EventType), but of a much more restrictive collection: ProblemSolvingType. Moreoever, it must be a spec of SolvingAProblem. Given the system's current lack of robotic sensors/effectors, it is pretty much restricted to tasks that (try to) carry out some sort of computational procedure. So, for example, FindingPrimeNumbers is something it might be able to do, or PlanningACircumnavigationRoute, but not CircumnavigatingTheGlobe (at least, not until it gets lots of additional peripherals!)." pg. 150 "Syntactically speaking, a *thing* is anything *about which* you can state a fact, and *to which* you can ascibe a name. Any *thing* can be represented by a unit in Cyc, and each Cyc unit (that is, frame) represents one *thing*. So extensionally, *unit* and *thing* are synonymous." pg. 356-358 "Cyc's Final Exam---Where We Hope To Be in Late 1994. In brief, the goal of our project is to break the brittleness bottleneck once and for all. Less briefly, we could state some "final exam" questions for Cyc in late 1994. No, we don't expect Cyc to be all-wise then, but we do hope to have moved by then from the manual 'brain-surgery' mode of knowledge entry, to a more 'Socratic dialogue' knowledge acquisition metaphor. These 'final exam' questions would be: 1a. Build an expert system in Cyc and show how it is *better* than building it from scratch (that is, in a conventional expert system building shell). ... 1b. Build two expert systems in Cyc, on related topics, neither projects knowing much about the other. Show that they can *share* each other's knowledge productively. 'Sharing' means that the whole is more than the sum of the parts---that the combined system does *better* than just the union of what the two individual systems could do. 'Better' includes such dimensions as: a wider range of problems can be handled, more correct answer are generated, fewer mistakes are made, the system is more easily understood by the user and/or it is faster. ... 2a. Communicate with Cyc in English, by dint of a natural language understanding program built largely by ACA's Human Interface project, which is deeply, recursively intertwined with Cyc. The understanding process includes disambiguation, resolving anaphora and ellipsis, as well as more sophisticated tasks such as recognizing and handling hyperbole, metaphor, and humor. After being understood, the new communication will then be added to Cyc's KB. If it was a question, then Cyc's answer will help guage its understanding of both the query and of its existing KB. ... 2b. Demonstrate that, based on 2a, the Cyc project is no longer needed. The massive hand-coding knowledge representation effort can be wound down and eliminated by the end of 1994. In its stead, Cyc will grow by assimilating textbooks, literature, newspapers, etc. It will also have access to, and models of, numerous large data bases to call on as needed. In place of the knowledge enterers will be a cadre of teachers, to elucidate confusing passages, to discuss and educate rather than to continue to practice brain surgery upon Cyc's KB. 3. Demonstrate that Cyc can learn by discovery. This goes beyond the sort of 'learning by being programmed' of 1a and 1b, and beyond the sort of 'learning by reading' of 2a and 2b. It includes deciding what data to gather, noticing patterns and regularities in the data, and drawing from those patterns useful new analogies, dependencies, and generalizations. This sort of learning may go on either *proactively* ('at night' or on idle machines) or *dynamically* as needed in the course of solving some tough problem. 4. Have Cyc be the major 'consensus reality KB' for the world. Dozens of projects, throughout academia and industry, make use of Cyc, build new knowledge into it, and send and receive constant updates to keep the KBs consistent. Cyc becomes thought of as *the knowledge utility*, much like an electric or telephone utility today: something everyone plugs into and is happy to pay for, Just as, today, no one would even *think* of buying a computer that didn't have an operating system and that couldn't run a spreadsheet and a word processing program, we hope that by 1999 no one would even *think* about having a computer that doesn't have Cyc running on it." ------------------------------ Date: Fri, 4 Jun 93 10:53:08 MDT From: ingres@cs.unm.edu Subject: TECH: DAK = trash Ronald Hale-Evans wrote: >As it turns out, I am not alone in observing that this unit is crap. >When I got it, I signed up on a shortwave mailing list, and many >people told me they had had the same bad experience with DAK shortwave >equipment. The DAK warranty is a ripoff in my opinion, so you won't >get any satisfaction from that. I don't know about this particular piece of equipment, but I long ago began throwing the DAK catalog away unread, because everything I bought from them was crap. (Statistics: 100% of three items + a very unhelpful phone complaint staff + 0 replies to 5 letters I sent them. They do have a very efficient billing and dunning staff.) Everyone I've talked to about them has had similar problems. I'm curious, is there anyone on this list who has bought something from DAK and been satisfied with the deal? Neil Hammar hammar@unmvax.cs.unm.edu ------------------------------ Date: Fri, 04 Jun 1993 13:08:39 -0400 From: "Perry E. Metzger" Subject: Power Crazed FDA X-Reposting-Policy: redistribute only with permission KMOSTA01@ulkyvx.louisville.edu says: > > Heath Goebel just does not understand -- you see, FDA really needs to > ban all life-protecting drugs, because they do not work, every person > on record who took them eventually died, except for some limited > number or survivors, and for those death seems likely! FDA must act now! Indeed, virtually every person who has ever breathed oxygen has died. Its time for immediate EPA action to ban this horribly reactive oxidizer that is polluting our atmosphere. .pm ------------------------------ Date: Fri, 04 Jun 1993 12:58:17 -0400 From: "Perry E. Metzger" Subject: FWD: To destroy an anarchy (fwd) X-Reposting-Policy: redistribute only with permission Nick Szabo says: > > There must be a government, or government will be imposed. So says > the net.wisdom on aleph & FutureC. Since few people in the world have > ever seen a culture without government in the 20th century, this could > become a very popular meme... any ideas for a counter-petition to > keep "The Net" an anarchy? None are needed. It is not possible to impose governance on the net. There is no longer any organization on earth that is in practice capable of altering its form. Its transnational, rapidly expanding, and has no central organization of any sort. Short of a treaty that all the major western powers agreed to to impose heavy regulation on every computer, it would be hard to do anything to it. .pm ------------------------------ Date: Fri, 04 Jun 1993 13:00:42 -0500 (EST) From: KMOSTA01@ULKYVX.LOUISVILLE.EDU Subject: Power Crazed FDA Heath Goebel just does not understand -- you see, FDA really needs to ban all life-protecting drugs, because they do not work, every person on record who took them eventually died, except for some limited number or survivors, and for those death seems likely! FDA must act now! Krzys' ------------------------------ Date: Fri, 04 Jun 1993 13:05:47 -0400 From: "Perry E. Metzger" Subject: TECH: shortwave X-Reposting-Policy: redistribute only with permission Ronald Hale-Evans says: > >Catalog item #5769 $79.90 + $6 P&H 800-DAK (325)-0800 Anytime > > > >Dual stereo cassettes (1 records, both play) 530-1630kHz (AM) > >Programmable timer 3200-7300kHz (Shortwave1) > >20 presets for tuner 9500-21750khz (Shortwave2) > >etc. 87.5-107.9MHz (FM/Stereo) > > WARNING!!! This equipment is a piece o' shit. DO NOT BUY IT. I made > that mistake, and within about forty hours of usage, the following > things went wrong with it: This brings up an interesting question for the list. I've been a fan of the BBC World Service's news program for some time, but unfortunately I can't get it anymore by rebroadcast on a local station. For a while, I've contemplated doing something neat like getting a shortwave, recording an overnight broadcast, and listening to the BBC in the morning while showering instead of National Socialist Radio. (The BBC World Service is one of the few government instititions on earth I legitimately like -- when all the governments on earth collapse some way must be found of privatizing it and making it turn a profit). Can anyone recommend a good and not too expensive shortwave system? I need something that would be capable of picking up the Beeb from the inside of a Manhattan apartment building. I can likely handle the time based recording myself -- at worst, I'll hook up the output to my home sparcstation and record digitally to disk. Perry ------------------------------ Date: Fri, 4 Jun 93 19:12:51 +0100 From: flumiani@udmi5400.cineca.it (Federico Flumiani) Subject: help Extropians Hello, Udine, 6-4-93 i am an italian student in computer science, interested in all new kind of things like VR, nanotechnology,the new frontiers of mankind, so i feel very excited when i saw your announce in the last number (2nd) of WIRED the mag. So why not to send a mail for getting infomation about this great movement? I will be very happy, if you will send me more news about EXTROPHY. Ciao , Federico #+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+#+# For Virtual replay my e-mail address is: flumiani@udmi5400.cineca.it ------------------------------ Date: Fri, 4 Jun 93 11:35:59 PDT From: Robin Hanson Subject: PHYS/COMP: Quantum Computer Feasible Now?! Amazing if true. I realize this is tantilizingly short, but it's all I have. Robin ------- Start of forwarded message ------- From: poc@im.lcs.mit.edu (Physics of Computation Seminar) To: physics.computation@hc.ti.com Message-Id: <9306032126.AA04685@im.lcs.mit.edu> Date: Thu, 3 Jun 93 17:26:36 EDT Subject: a new seminar series Return-Path: Note: Although this seminar series will of course be of most interest to people in the Boston area, others may wish to subscribe to receive seminar postings just to see the abstracts of the talks. **************************************************************************** **** NEW SEMINAR SERIES **** NEW SEMINAR SERIES **** NEW SEMINAR SERIES **** **************************************************************************** PHYSICS OF COMPUTATION SEMINAR Date: Monday, June 7 Time: 2PM Place: Massachusetts Institute of Technology Room: 6-120 A Technologically Feasible Quantum Computer Seth Lloyd Complex Systems Group & Center for Nonlinear Studies Los Alamos National Laboratory Abstract: Arrays of weakly-coupled quantum systems can be made to compute by subjecting them to a sequence of electromagnetic pulses of well-defined frequency and length. Such pulsed arrays are true quantum computers: bits can be placed in superpositions of 0 and 1, logical operations take place coherently, and dissipation is required only for error correction. When operated at optical frequencies, the array could provide massively parallel computation at clock cycles of pico- to nano- seconds. None of the steps involved in constructing or operating such a computer lie out of the reach of current technology. Host: Norm Margolus, MIT Lab for Computer Science This is the first talk in what will be an occasional series of seminars on "The Physics of Computation." Our goal in initiating this series is to provide a forum for a community of researchers that otherwise wouldn't have a natural venue. The stress will be on fundamental physical issues that govern efficient "engines" that transform information, and on the role of information and computation in understanding natural systems. For example, can we characterize the most energy-efficient class of information engines? Can atomic-scale devices compute? Can "quantum computers" do anything that ordinary computers can't? Can an arbitrary quantum state be broken up into fixed-sized quantum signals and transmitted? We will also include discussions on the usefulness of novel architectures that are recommended by fundamental physical constraints. For example, the finite speed of light is a constraint that favors machines with only short wires: What kinds of computations are well adapted to such machines? How much is known about particular classes of computations, such as Lattice Gases, that answer this question? What are the outstanding questions about the capabilities of such models? Please forward this notice to anyone who you think might be interested. This initial mailing will be sent to several mailing lists; if you wish to receive subsequent notices please reply to "poc@im.lcs.mit.edu". Suggestions for speakers (particularly if an interesting visitor is going to be passing through the Boston area) should also be sent to this address. This series is being sponsored by the MIT Information Mechanics Group (Lab for Computer Science), in conjunction with the MIT Physics and Media Group (Media Lab), the MIT Porous Flow Project (Earth, Atmospheric and Planetary Sciences), and the Mathematical Sciences Research Group at Thinking Machines Corporation. ------- End of forwarded message ------- ------------------------------ Date: Fri, 4 Jun 93 12:44:53 PST From: "Mark W. McFadden" Subject: SPACE: Fermi paradox, etc. >> Why do all these nanotechnology/wormhole using races all seem >> to breed like bunnies? > >Why to bunnies breed? Bugs? Flies? Why does every species breed? >Darwin. Some. of course, may go the way of the dodo, but then we don't >care about them. It the ones that don't that nuke us. > Why do you always assume I don't _understand_ when I don't agree with your simplistic conclusions? Okay, Darwin time as per Metzger et al. The fittest organism will outreproduce less fit (suitable) organisms. This is because they are better adapted for their environment, so with all those survival advantages they will reproduce more often through their longer lives, and more of their offspring will survive because they eat better, and the ones carrying the fitter genes will over time and through generations come to dominate and they will (cuz of Darwin) just keep multiplying because that's the goal of evolution, lots and lots of whatever's successful. So, all we have to do to see what the best, fittest survival tactic is for the human species, is look around and see who's out-reproducing everyone else. I guess we can expect some great things coming out of Calcutta, and mainland China and Rio real soon now. Or does Darwin become superfluous when discussing an intelligent species? _____________________________________________________________________ | Mark W. McFadden | Been there....done that. mwm@wwtc.timeplex.com | __________________________________|__________________________________ ------------------------------ Date: Fri, 04 Jun 1993 15:39:44 -0400 From: "Perry E. Metzger" Subject: SPACE: Fermi paradox, etc. X-Reposting-Policy: redistribute only with permission "Mark W. McFadden" says: > > >> Why do all these nanotechnology/wormhole using races all seem > >> to breed like bunnies? > > > >Why to bunnies breed? Bugs? Flies? Why does every species breed? > >Darwin. Some. of course, may go the way of the dodo, but then we don't > >care about them. It the ones that don't that nuke us. > > > > Why do you always assume I don't _understand_ when I don't agree with your > simplistic conclusions? Maybe because you don't show signs of understanding. We can't operate on psychic powers -- we are limited to judging you based on what you say. I don't think that its unreasonable to conclude from what you've said that you don't understand the arguments. > So, all we have to do to see what the best, fittest survival tactic is > for the human species, is look around and see who's out-reproducing > everyone else. > I guess we can expect some great things coming out of Calcutta, and > mainland China and Rio real soon now. Well, actually, I do expect that were the human race qua human race to continue on current trends for millenia to come, the vast bulk of the world would have Indian and African features, yes. (China has reversed its population trend of late by force.) In that sense, one can indeed assume that what they are doing makes a degree of sense, evolutionarily speaking. Whether you think that India is a great place to live is irrelevant. Of course, the reason for the rapid reproduction in these countries is that humans have a tendancy to switch survival strategies from quality of children to quantity when they are under environmental stresses -- better to have several mediocre children survive a brood of ten than have the the three you put lots of effort into die. As a long run strategy, the one "we" (not meaning caucasians but meaning folks in developed countries) are following will lead to more offspring -- because ours have a far smaller chance of starving to death or dying of plague and have tremendous resources available. Having lots of children is an emergency tactic. > Or does Darwin become superfluous when discussing an intelligent species? No, Darwin is still relevant, but you don't understand the issues. Perry ------------------------------ Date: Fri, 04 Jun 93 20:32:12 GMT From: price@price.demon.co.uk (Michael Clive Price) Subject: TECH: shortwave Perry Metzger: > The BBC World Service is one of the few government > institutions on earth I legitimately like I second that. I often wake up early in the morning/late night and lie tucked up in bed listening to it. It's great! So incredibly twee and informative. > -- when all the governments > on earth collapse some way must be found of privatizing it and making > it turn a profit). Okay, when we're all unspeakably nanotech-rich we'll form a (slave-run) charity just to broadcast World Service style. > Perry Mike Price price@price.demon.co.uk AS member (21/3/93) ------------------------------ Date: Fri, 4 Jun 1993 15:37:56 -0400 (EDT) From: Carol Moore Subject: PREFIXES: The 'README' meme Then use READ ME as a Prefix with Description like: READ ME: Send Me the Unsubscribe Address READ ME: Please Use Descriptive Subject Lines READ ME: Stop Flaming People or whatever.... On Thu, 3 Jun 1993, FutureNerd Steve Witham wrote: > Has anyone noticed that the name "README" is actually a very successful meme? > > Obviously there are practical reasons to have readme files, and reasons to > settle on a conventional name for such files. But the thing about the > name "README" is that it's imperative, and actually insists that you > pay attention to it. >. . . . . . (-: Like My Short Sig-Line?? -- cmoore@cap.gwu.edu :-) ------------------------------ Date: Fri, 4 Jun 93 14:31:58 CDT From: derek@cs.wisc.edu (Derek Zahn) Subject: SPACE: Fermi paradox, etc. Mark McFadden writes: > So, all we have to do to see what the best, fittest survival tactic is > for the human species, is look around and see who's out-reproducing > everyone else. > I guess we can expect some great things coming out of Calcutta, and > mainland China and Rio real soon now. > > Or does Darwin become superfluous when discussing an intelligent species? Only to the extent that intelligence reduces competition for resources, which it may do to a little but so far not much. I'm afraid I still don't understand your point about overpopulated areas. I suppose you want to say that the propensity for a population to breed itself to starvation is bad, but so what? Even if it's granted that rationality might tend to reduce the breeding rate in the face of scarce resources, I thought the discussion was about species behavior when expansion to new resources is possible. In this case, whatever minority has the propensity to so expand will do so, producing offspring which will also tend to do so. derek ------------------------------ Date: Fri, 4 Jun 1993 13:27:25 -0700 (PDT) From: szabo@techbook.com (Nick Szabo) Subject: EXTROPRENEUR: Request for Comments I find both the speech recognition (SR) & scanned-document processing (SDP) markets promising, so I can't help Derek decide on which, but I'll throw in some comments. These technology areas seem ideal for the "product wave" strategy of marketing. This has been one of the unsung strategies of Japanese consumer electronics: introduce dozens of slightly varying products, market them slightly differently (targeting different niches), and quickly withdraw the failures, leaving behind a still substantial number of products highly fit for the market. Don't try to fully anticipate consumer needs; instead anticipate a variety of reasonable needs and let the market pick the winner. Cypress Semiconductor also seems to do this; CEO T.J. Rodger's article in this months' _Reason_ claims they have only 150 "product designers" and 70 "technologists", but sell more than 1,500 products! Intel did this with CMOS in the early 70's; I'd love to hear Tim May's comments about early Intel marketing and how they made so many successful products with such a small effort in those days. For a startup, the product wave starts with a few narrowly targeted products, then when those start bringing in revenue use that money to open up the floodgates, letting loose a barrage of closely related products. The same small code library and marketing material can be mutated, crossed-over, and otherwise edited into dozens of forms until the market finds the best one. It's amazing how many companies become satisfied with the first flush of victory, and religiously stick to the initial, quite sub-optimal great idea. Witness Visi-Calc, DBase, Osborne, etc. Then too, many large companies throw vast sums into single products instead of produce waves. Startups cannot afford either mistake. The product wave should work even better for software or sofware-heavy electronics in SR & SDP. Given a debugged, good performance, and modular design library, making slight variations on design and releasing several products is relatively cheap. Unless the allegedly terrifying giant competitors like Apple use the product-wave strategy, they are bound to leave open hundreds of market niches, and probably the best ones at that. They can't pick the winners any better than you can, and they're used to selling all-purposes MacIntoshes, not niche-targeted products. In particular, I don't think they'll put the "I can't program my VCR" guy out of business unless they focus on that narrow niche at the expense of the thousands of other uses for SR. Derek has some great ideas for market niches; I especially find the scanning of addresses from business letters a promising area. Here are some more SDP niches, each one of which might support dozens of slightly different products: * scan in business forms, convert them to on-line forms * scan in legal contracts (for style checking, archiving, etc.) * scan in old blueprints -> CAD files * scan in prescription, patient status, etc. forms (including doctor's signatures :-) One targeted SDP product has already made millions for a startup: the Resumix (sp?) system which specializes in scanning in resumes for harried "Human Resource" departments. Here's a more general device I'd be interested in: a hand-held copy machine, basically a scanner hooked to a small portable computer. Take it into the library, use it to scan in the pages of a scientific article, which can then be taken home, uploaded to computer and printed out on laser printer or posted on the net. This wouldn't make publishers too happy, but if it fits in a briefcase tough luck for them. :-) In SR, consider what kinds of software would profit from voice input. Can we talk in C software code and editing commands, instead of typing? We all know how to pronounce "!" :-), but how does one pronounce "{", "(", etc.? It seems that most computer languages were designed with the keyboard, not the human voice in mind. On the other hand, we have the Q&A database query language, which was designed to be user-friendly by giving it a large, redundant vocabulary of English words. I've only looked through the manual, but its vocabulary looks awkward to type, and a large number of keystroke-shortcuts would be confusing. Q&A query language looks ideal for hooking up to special-purpose speech recognition. Even better might be a Q&A "clone" plus voice input to go over the top of more popular database packages like DBase or Oracle. Also consider the following: * Providing software access for people with two hands occupied at a machine, eg a car, sewing machine, lathe, etc. or handicapped. Voice-operated maps on CD-ROM is one of a vast number of possibilities here. * Computer games that respond to screams, groans, etc. * Music composition by humming or whistling * Special-purpose telephone applications (replace those dratted "punch 1 now" menus!) * Voice-input cash registers. ("Big Mac with Fries!" sufficient to both inform the cook & ring up the sale) In general, SR market research should look at: * under what conditions it's socially acceptable to use voice input (telephone companies have a head-start here) * what products are most suitable to being converted to voice (eg computer programs like Q&A with large command vocabularies) * what new things are made possible by voice input, eg people with two hands occupied at a machine can now simultaneously use use a computer But don't go with any of these ideas just because they sound good. Buy a list of small businesses in the market niche(s) of interest who've bought OCR scanners or general-purpose SR in the last couple years and call them up, asking what they've used them for, or what they tried to use them for but couldn't make to work. Talk to these people, don't pretend you more than they do about what they want. Then hire some of these people to help you design & sell the product. Nick Szabo szabo@techbook.com ------------------------------ End of Extropians Digest V93 Issue #0299 ****************************************