Diplomacy Zine -- EP #199 Chapter Seven From: Eric_S_Klien@cup.portal.com Date: Thu, 20 Sep 1990 00:02:21 +0000 Issue #199 of ELECTRONIC PROTOCOL: ************************************************************************ BART (MUMBLING) Lousy, washed up, broken down... old tub of guts... who does... HOMER What's the matter, boy? BART He wouldn't sign my ball. MARGE Well, I thought those ballplayers were supposed to set an example for children. Bart, give me that ball! Bart flips her the ball and, with great resolve, she marches down the aisle to the field. ************************************************************************ Chapter One contains: BLITZKRIEG, GETTYSBURG, RED STORM RISING, and PASSCHENDAELE And is published by daybell@aludra.usc.edu/Donald Daybell Chapter Two contains: DRAGONSLAYER, JACAL, MANHATTAN, VERSAILLES, DRESDEN, and KHAN And is published by sinhaa@mcmaster.ca/Anand Sinha Chapter Three contains: DAWN PATROL, BERLIN, EL ALAMEIN, SQUALANE, UNGAWE, CAPTAIN CAVEMAN, BRUSILOV OFFENSIVE And is published by cwekx@htikub5.bitnet/Constantijn Wekx Chapter Four contains: NICKEL, OZARK, DEADLY DAGGERS, YORKTOWN, MONTREUIL-SUR-MER, FIRE WHEN READY Chapter Five contains: ARCHANGEL, BORDEL, ERIS, MASADA, and YALTA And is published by jjcarette@watami.waterloo.edu/David Gibbs Chapter Six contains: TOKUGAWA, BERLIN WALL, HIROSHIMA, GENGHIS KHAN, SEA LION, VIOLENT PEACE And is published by ps9zrhmc@miamiu.bitnet/Peter Sweeney ------------ Chapter Seven ------------ No games in this issue. Publisher comments: Scene is from "Dancin' Homer", a Simpsons episode that has not been released yet. I'll publish a little more in my next issue. By the way, I'll be sending in my Simpsons script for WGA registration within the next day or so, and then I'll submit it to Gracie Films. Wish me luck! From August 9, 1990 Investor's Daily: MAIL BOXES ETC. STRENGTHENS ITS LINKS WITH OTHER SERVICES By John A. Jones The linkup announced a month ago between Mail Boxes Etc. and United Parcel Service marks a new step forward for Mail Boxes, the leading alternative to the struggling U.S. Postal Service. Mail boxes, founded 10 years ago in San Diego, is the nation's leading franchiser of neighborhood service centers for postal and communications services for individuals and small businesses. A year ago, it opened its 1,000th franchise, and it's adding about one a day. The network spans some 1,300 offices in neighborhood shopping centers, with plans to reach 2,500 in five years. Centers are open in 45 states plus Puerto Rico, Canada, Mexico and Japan. The mailbox service acts something like a retail agent for the Postal Service, which still carries all the regular mail collected and distributed through private centers. But the private boxes can have "suite" numbers in their addresses, they're accessible 24 hours a day, and customers can phone in to see whether they have any mail. FROM STAMPS TO FAXES Besides selling stamps, envelopes and other routine postal items, Mail Boxes centers act as outlets for DHL Airways Inc., UPS and other private carriers; Western Union communications and financial services, facsimile machine and telex, phone message and voice mail services; and for computer terminals. They also sell office supplies and services, including document copying and word processing. Mail Boxes has strengthed its ties with three other leaders in the private communication business this year. Last spring, it signed a two-year agreement for its centers to become Western Union money transfer agents, as well as an exclusive pact with DHL Airways as its preferred air express shipper. Mail Boxes franchises get a preferred volume rate when their centers become designated drop-off points for DHL. Mail Boxes announced an alliance with UPS on July 9, through a letter of intent to sell UPS 400,000 shares of MBE common stock for $10.1 million, plus warrants for 400,000 more shares for $1.2 million. If the warrants are exercised, UPS would own about 17% of Mail Boxes. The agreement also would allow UPS to increase its equity in MBE to 33% after one year, and to name a director to MBE's board. Per-share earnings for the fiscal fourth quarter ended April 30 rose 33% to 24 cents from 18 cents a year earlier. Net income was up 38% to $927,000 from $671,000. Revenue gained 20% to $6.48 million from $5.41 million. For the full fiscal year, net income fell 37% to $1.52 million, or 40 cents a share, from $2.39 million, or 63 cents a share, the previous year. Net was reduced by $1.7 million, or 45 cents a share, after the settlement of a lawsuit by Nancy Tash, daughter of one of the founders, over the value of stock she sold the company in 1983. Operating income was up 47% to $4.75 million from $3.24 million. Revenue rose 33% to $25.7 million from $19.4 million. President and Chief Executive A. W. (Tony) DeSio, who co-founded the company and has headed it sine 1983, said Mail Boxes reached a vital goal a few months ahead of schedule last August, when its network reached 1,000 U.S. centers. The company also has master licenses in place to extend the franchise system throughout Canada and Japan, and a joint venture to run a pilot center in Mexicali, just across the Mexican border south of California. The Mexican company is intrested in a master franchise agreement for Mexico, DeSio told Investor's Daily, but it's waiting to make sure the Mexican government won't restrict its operations. PLANS FOR EUROPE "We hope to nail down at least one master license this year in Europe", DeSio said. The European Community's unification in 1992 "opens the door for a multinational European system for communications" when government regulations catch up with economic developments there, he said. DeSio said that as the network grows nationwide, larger companies are asking Mail Boxes to act as a national retail distribution system for them. For example, he said, Mail Boxes franchises now can offer direct electronic income-tax filing service, and access to rented mailing lists. "They are far and away the leader" in the private mail service market, said Roger Lipton, an analyst at Ladenburg, Thalmann & Co. in New York. Mail Boxes locations exceed 1,300, he said, "and I'm not aware of anybody else with more than a couple of hundred". Lipton noted that MBE's balance sheet is strong, with $2.3 million in cash and no long-term debt. I received the following from CompuServe: Here is an on-line version of the Runestone Poll results packet which was mailed to everyone who sent in 50 cents or two U.S. 25 cent stamps: ----------------------------------------------------------------------- ********************************* * * * 1990 Runestone Poll Results * * * ********************************* July 1, 1990 Dear Postal Gamer, Polling for the 1990 Runestone Poll is finished; here are the results. IUd like to thank everyone who voted -Q this year I received exactly 250 ballots. Thanks too to those who helped publicize the Poll by distributing ballots, and to all of you who sent me a donation to help cover expenses. As usual, I enclose the three main lists in their entirety -Q those for zines, subzines, and GMs. This year, however, I also include a one-page Poll summary which lists only those main list entries which finished at or above average. If youUre a publisher, IUd like you to consider publishing the one-page summary rather than the whole of the main lists; this may be a way to avoid focusing attention on those who finished near the bottom. Let me remind you all again don't take the Poll too seriously. This is a hobby, and our goal is to have fun. If you finished near the top, congratulations! But if your ranking is lower than youUd like, donUt be too upset. IUd be delighted if you make some improvements as a result, but if you and your readers enjoy what youUre doing donUt stop just because of what happened in the Poll. Let me add that the Poll publication, which will include complete stats and a number of articles, should be out sometime in August. If you havenUt ordered it yet, you may do so by sending me $5.00. Sincerely, Eric Brosius 41 Hayward St. Milford, MA 01757 USA ----------------------------------------------------------------------- ********************************* * * * The 1990 Runestone Poll * * * ********************************* Main List entries finishing at or above average *** Zines *** (75 on main list) Rk Zine Name Votes Score 1 Northern Flame 40 8.716 2 Upstart 36 8.607 3 The Zine Register 72 8.488 4 Perelandra 31 8.392 5 Penguin Dip 30 8.310 6 Kathy's Korner 35 8.284 7 Been There, Done That 76 8.228 8 House of Lords 43 7.986 9 The Boob Report 22 7.808 10 ark 12 7.736 11 White House Mania 20 7.694 12 The Metadiplomat 25 7.674 13 Carolina Cmd & Cmntry 57 7.642 14 Excelsior 41 7.505 15 benzene 36 7.469 16 The Armchair Diplomat 15 7.452 17 Buckeye Rail Gazette 14 7.217 18 Moire 36 7.184 19 Cheesecake 27 7.169 20 Graustark 15 7.134 21 Maniac's Paradise 18 7.133 22 The Canadian Diplomat 49 7.129 23 Passchendaele 38 7.058 24 Fol Si Fie 14 7.042 25 Fiat Bellum 32 6.815 26 Dipadeedoodah! 46 6.766 27 Hagalil Hamaarvi 38 6.735 28 TRAX 10 6.689 29 Comrades in Arms 40 6.668 30 A Sharp Mind 19 6.621 31 Bushwacker 55 6.514 32 Ohio Acres 21 6.487 33 The Abyssinian Prince 16 6.441 34 Megalomaniac 22 6.429 35 Rebel 59 6.336 36 Entropy 27 6.277 37 Angst United 14 6.269 38 Pilot Light 18 6.189 *** Subzines *** (30 on main list) Rk Subzine Name Votes Score 1 High Inertia 41 8.819 2 The Eccentric Diplomat 8 8.545 3 The Unabashed Bo(t) 5 8.529 4 Extremism in Defense... 20 7.596 5 Acropolis 13 7.394 6 Shut Up Jack!!! 21 6.948 7 Standard Deviation 8 6.887 8 Poll Talk 7 6.886 9 Asterion 12 6.803 10 CDO News 23 6.552 11 Tennessee Rails 21 6.365 12 Notes from the Bunker 21 6.206 13 KarmelUs Panorama 23 5.946 14 War Correspondant 9 5.861 15 McKee Raves 18 5.656 *** GMs *** (48 on main list) Rk GM Name Votes Score 1 Russ Blau 6 9.022 2 Kathy Caruso 11 8.814 3 Andy Lischett 14 8.614 4 Randolph Smyth 5 8.413 5 Jim Benes 6 8.364 6 Jim Burgess 8 8.340 7 Garret Schenck 15 8.222 8 Fred Davis 17 7.791 9 Vince Lutterbie 10 7.699 10 Francois Cuerrier 9 7.677 11 Douglas Kent 7 7.593 12 Fred Hyatt 14 7.545 13 David Hood 9 7.461 14 Robert Cochran 5 7.426 15 Pete Gaughan 14 7.414 16 Ken Hill 5 7.413 17 Stephen Dorneman 11 7.397 18 Tom Swider 7 7.238 19 Phil Reynolds 13 7.235 20 Eric Klien 7 7.084 21 Marc Peters 9 7.056 22 Bob Given 7 6.902 23 Tim Moore 8 6.896 24 George Mann 7 6.584 ----------------------------------------------------------------------- ********************************* * * * The 1990 Zine Poll * * * ********************************* Rk %ile Name Votes Percentile Final Modified Pref Change Score mean score 1 [100] Northern Flame 40 Rose by 11 8.716 8.344 9.459 2 [ 99] Upstart 36 Rose by 25 8.607 8.451 8.919 3 [ 97] The Zine Register 72 Rose by 11 8.488 8.069 9.324 4 [ 96] Perelandra 31 Fell by 3 8.392 7.960 9.257 5 [ 95] Penguin Dip 30 No change 8.310 8.208 8.514 6 [ 93] Kathy's Korner 35 Rose by 33 8.284 8.000 8.851 7 [ 92] Been There, Done That 76 New to list 8.228 7.714 9.257 8 [ 91] House of Lords 43 No change 7.986 7.689 8.581 9 [ 89] The Boob Report 22 Rose by 40 7.808 7.557 8.311 10 [ 88] ark 12 New to list 7.736 8.125 6.959 11 [ 86] White House Mania 20 New to list 7.694 8.500 6.081 12 [ 85] The Metadiplomat 25 New to list 7.674 7.525 7.973 13 [ 84] Carolina Cmd & Commentary 57 Fell by 13 7.642 7.511 7.905 14 [ 82] Excelsior 41 Rose by 13 7.505 7.338 7.838 15 [ 81] benzene 36 Fell by 15 7.469 7.556 7.297 16 [ 80] The Armchair Diplomat 15 New to list 7.452 7.833 6.689 17 [ 78] Buckeye Rail Gazette 14 New to list 7.217 8.089 5.473 18 [ 77] Moire 36 New to list 7.184 7.229 7.095 19 [ 76] Cheesecake 27 Rose by 1 7.169 7.139 7.230 20 [ 74] Graustark 15 Fell by 3 7.134 7.458 6.486 21 [ 73] Maniac's Paradise 18 New to list 7.133 7.153 7.095 22 [ 72] The Canadian Diplomat 49 Fell by 13 7.129 7.383 6.622 23 [ 70] Passchendaele 38 Fell by 23 7.058 7.007 7.162 24 [ 69] Fol Si Fie 14 Fell by 3 7.042 7.589 5.946 25 [ 68] Fiat Bellum 32 Fell by 32 6.815 6.742 6.959 26 [ 66] Dipadeedoodah! 46 Rose by 14 6.766 7.109 6.081 27 [ 65] Hagalil Hamaarvi 38 Fell by 6 6.735 7.230 5.743 28 [ 64] TRAX 10 Fell by 20 6.689 7.500 5.068 29 [ 62] Comrades in Arms 40 Fell by 29 6.668 7.062 5.878 30 [ 61] A Sharp Mind... 19 Rose by 19 6.621 7.059 5.743 31 [ 59] Bushwacker 55 Rose by 2 6.514 7.068 5.405 32 [ 58] Ohio Acres 21 Rose by 20 6.487 6.994 5.473 33 [ 57] The Abyssinian Prince 16 New to list 6.441 7.094 5.135 34 [ 55] Megalomaniac 22 New to list 6.429 6.670 5.946 35 [ 54] Rebel 59 Fell by 28 6.336 6.936 5.135 36 [ 53] Entropy 27 New to list 6.277 6.477 5.878 37 [ 51] Angst United 14 New to list 6.269 6.768 5.270 38 [ 50] Pilot Light 18 New to list 6.189 7.222 4.122 39 [ 49] Not New York 20 Fell by 27 6.131 7.000 4.392 40 [ 47] Disease City 15 Fell by 33 6.113 7.750 2.838 41 [ 46] Diplomag 16 Rose by 25 5.895 7.188 3.311 42 [ 45] Electronic Protocol 72 Fell by 23 5.848 7.590 2.365 43 [ 43] Cathy's Ramblings 37 Fell by 2 5.831 6.416 4.662 44 [ 42] Ter-ran 18 Rose by 15 5.829 6.278 4.932 45 [ 41] Vertigo 34 Rose by 4 5.805 6.309 4.797 46 [ 39] Son of Flip 16 Fell by 9 5.794 6.359 4.662 47 [ 38] Get Them Dots Now! 33 Fell by 16 5.790 6.152 5.068 48 [ 36] The Appalachian General 29 Fell by 23 5.674 6.349 4.324 49 [ 35] Diplomacy Digest 38 Fell by 21 5.601 6.138 4.527 50 [ 34] Mondoj 11 New to list 5.599 6.000 4.797 51 [ 32] Tyromania 11 Rose by 17 5.585 6.216 4.324 52 [ 31] Clandestine Activities 30 Fell by 24 5.572 6.500 3.716 53 [ 30] Dippy 13 Fell by 34 5.549 6.769 3.108 54 [ 28] Blade Wars 10 New to list 5.392 6.500 3.176 55 [ 27] Retaliation 23 Fell by 39 5.388 5.582 5.000 56 [ 26] Everything... 34 Fell by 7 5.376 6.206 3.716 57 [ 24] Perestroika 29 New to list 5.269 6.181 3.446 58 [ 23] The Messenger 10 Fell by 16 5.239 6.000 3.716 59 [ 22] Down at the Mouth 22 New to list 5.186 5.989 3.581 60 [ 20] The Home Office 33 Fell by 23 5.162 6.189 3.108 61 [ 19] Ouinipique 12 Fell by 6 5.060 5.833 3.514 62 [ 18] So I Lied 25 Rose by 1 4.969 5.900 3.108 63 [ 16] Diplomacy World 47 Fell by 34 4.797 5.912 2.568 64 [ 15] Protocol 21 Fell by 15 4.744 6.238 1.757 65 [ 14] Diplomacy Tribune 14 New to list 4.451 5.696 1.959 66 [ 12] Boast 18 Rose by 2 3.940 5.167 1.486 67 [ 11] Dark Mirror 30 Fell by 77 3.884 5.083 1.486 68 [ 9] Alpha & Omega 24 Fell by 2 3.725 4.844 1.486 69 [ 8] Costaguana 14 Fell by 71 3.639 4.411 2.095 70 [ 7] The Megadiplomat 25 Fell by 13 3.552 4.450 1.757 71 [ 5] The Assassin's Blade 15 New to list 3.081 4.250 0.743 72 [ 4] When The Lights... 19 New to list 3.035 3.809 1.486 73 [ 3] Hansard 13 Fell by 11 2.876 3.808 1.014 74 [ 1] Known Game Openings 38 Fell by 1 2.603 3.533 0.743 75 [ 0] KGO Zn Drct (Carrier) 22 New to list 1.879 2.784 0.068 ----------------------------------------------------------------------- ********************************* * * * The 1990 Subzine Poll * * * ********************************* Rk %ile Name Votes Percentile Final Modified Pref Change Score mean score 1 [100] High Inertia 41 No change 8.819 8.488 9.483 2 [ 97] The Eccentric Diplomat 8 New to list 8.545 8.938 7.759 3 [ 93] The Unabashed Bo(t) 5 New to list 8.529 9.000 7.586 4 [ 90] Extremism in Defense... 20 Rose by 5 7.596 7.688 7.414 5 [ 86] Acropolis 13 New to list 7.394 7.385 7.414 6 [ 83] Shut Up Jack!!! 21 New to list 6.948 7.060 6.724 7 [ 79] Standard Deviation 8 Rose by 25 6.887 6.969 6.724 8 [ 76] Poll Talk 7 New to list 6.886 7.054 6.552 9 [ 72] Asterion 12 New to list 6.803 6.583 7.241 10 [ 69] CDO News 23 Rose by 4 6.552 6.380 6.897 11 [ 66] Tennessee Rails 21 New to list 6.365 6.530 6.034 12 [ 62] Notes from the Bunker 21 Fell by 11 6.206 6.119 6.379 13 [ 59] Karmel's Panorama 23 Rose by 1 5.946 6.418 5.000 14 [ 55] War Correspondant 9 New to list 5.861 5.861 5.862 15 [ 52] McKee Raves 18 New to list 5.656 5.639 5.690 16 [ 48] The Melnibone Herald 12 Rose by 25 5.401 6.292 3.621 17 [ 45] The Whipping Post 10 Fell by 1 5.270 5.750 4.310 18 [ 41] Subwithnoname 12 Rose by 11 5.268 5.833 4.138 19 [ 38] Magus 17 Fell by 24 5.230 5.949 3.793 20 [ 34] Water on the Knee 16 New to list 5.208 5.484 4.655 21 [ 31] Nash Rants 12 New to list 4.862 5.396 3.793 22 [ 28] Foot In Mouth 44 Rose by 24 4.481 5.256 2.931 23 [ 24] Top Knife 6 Rose by 16 4.463 4.625 4.138 24 [ 21] The New Utopia 10 Fell by 29 4.356 5.500 2.069 25 [ 17] Horsin' Around 11 New to list 4.169 4.443 3.621 26 [ 14] Ring Finger in Rear 21 New to list 3.702 4.690 1.724 27 [ 10] Obfuscate 16 New to list 3.659 4.281 2.414 28 [ 7] Reginald Maudlin... 7 New to list 3.612 4.643 1.552 29 [ 3] OPERABLE 33 Rose by 3 3.243 3.572 2.586 30 [ 0] The First Negotiator 10 Fell by 19 2.549 2.875 1.897 ----------------------------------------------------------------------- ********************************* * * * The 1990 GM Poll * * * ********************************* Rk %ile Name Votes Percentile Final Modified Pref Change Score mean score 1 [100] Russ Blau 6 New to list 9.022 9.917 7.234 2 [ 98] Kathy Caruso 11 Rose by 10 8.814 9.125 8.191 3 [ 96] Andy Lischett 14 Rose by 2 8.614 9.357 7.128 4 [ 94] Randolph Smyth 5 New to list 8.413 9.375 6.489 5 [ 91] Jim Benes 6 Rose by 1 8.364 9.833 5.426 6 [ 89] Jim Burgess 8 Rose by 20 8.340 9.000 7.021 7 [ 87] Garret Schenck 15 New to list 8.222 8.875 6.915 8 [ 85] Fred Davis 17 Rose by 18 7.791 8.441 6.489 9 [ 83] Vince Lutterbie 10 New to list 7.699 8.250 6.596 10 [ 81] Francois Cuerrier 9 Fell by 16 7.677 8.431 6.170 11 [ 79] Douglas Kent 7 New to list 7.593 8.464 5.851 12 [ 77] Fred Hyatt 14 Rose by 28 7.545 8.232 6.170 13 [ 74] David Hood 9 Fell by 10 7.461 8.000 6.383 14 [ 72] Robert Cochran 5 New to list 7.426 8.000 6.277 15 [ 70] Pete Gaughan 14 Fell by 12 7.414 8.089 6.064 16 [ 68] Ken Hill 5 New to list 7.413 7.875 6.489 17 [ 66] Stephen Dorneman 11 New to list 7.397 8.330 5.532 18 [ 64] Tom Swider 7 New to list 7.238 8.357 5.000 19 [ 62] Phil Reynolds 13 New to list 7.235 8.087 5.532 20 [ 60] Eric Klien 7 Fell by 16 7.084 8.232 4.787 21 [ 57] Marc Peters 9 New to list 7.056 8.083 5.000 22 [ 55] Bob Given 7 New to list 6.902 7.375 5.957 23 [ 53] Tim Moore 8 New to list 6.896 7.844 5.000 24 [ 51] George Mann 7 New to list 6.584 7.482 4.787 25 [ 49] Don Del Grande 7 New to list 6.548 7.375 4.894 26 [ 47] Lee Kendter Jr. 8 New to list 6.546 7.531 4.574 27 [ 45] Tom Nash 16 New to list 6.543 7.688 4.255 28 [ 43] Steve Heinowski 11 Rose by 21 6.523 7.125 5.319 29 [ 40] Ran Ben-Israel 14 Fell by 14 6.442 7.482 4.362 30 [ 38] Constantijn Wekx 6 New to list 6.384 6.917 5.319 31 [ 36] Rob Greier 10 Fell by 3 6.348 7.500 4.043 32 [ 34] Jeff McKee 14 New to list 6.300 7.536 3.830 33 [ 32] Cal White 14 Fell by 1 6.288 7.411 4.043 34 [ 30] Brad Wilson 8 New to list 6.248 7.031 4.681 35 [ 28] Jason Bergmann 8 New to list 6.133 7.125 4.149 36 [ 26] Bob Acheson 20 Fell by 53 6.122 7.375 3.617 37 [ 23] Melinda Holley 27 Fell by 37 6.010 7.046 3.936 38 [ 21] Paul Gardner 7 Fell by 51 5.955 7.018 3.830 39 [ 19] Doug Acheson 12 Fell by 38 5.906 7.104 3.511 40 [ 17] Dave McCrumb 15 New to list 5.894 7.458 2.766 41 [ 15] Richard Kirby 5 New to list 5.715 6.125 4.894 42 [ 13] Don Williams 11 Fell by 39 5.514 6.568 3.404 43 [ 11] Dick Martin 12 Fell by 35 5.398 5.917 4.362 44 [ 9] Conrad von Metzke 12 Fell by 1 5.271 6.896 2.021 45 [ 6] Fred Wiedemeyer 5 Fell by 18 5.099 6.000 3.298 46 [ 4] Bruce McIntyre 8 Rose by 1 4.301 5.281 2.340 47 [ 2] Audrey Jaxon 7 New to list 3.257 3.875 2.021 48 [ 0] Simon Billenness 5 Fell by 15 1.848 0.750 4.0 From August 1990 Discover: COMMON SENSE AND THE COMPUTER By David H. Freedman Karen Pittman is hunched over a copy of How to Troubleshoot and Repair Any Small Gas Engine. It is not a gripping read. Pittman doesn't care much about the subject, and when she finishes she will probably be no more able to repair a balky lawn mower or outboard motor than she was when she began. What she's looking for in the arcane volume is something subtler: the bits of everyday knowledge that the authors didn't bother to explain to the reader at all -- what a lawn is, for instance, or where you're likely to find one. Every complex repair rule in the book may be predicated on dozens of such commonsense scraps, tidbits of general knowledge that are in need of no explanation. After Pittman collects them all, she must explain every one to a computer. At the end of the session, the computer still won't know a fraction of what a mechanic knows about small engines. But it will know everything Pittman knows about them. The computer's data base is filled with such mundane stuff: what to do with a knife and fork, what cats and dogs are, where the state of Texas is. There's a little art and biology and history thrown in, but nothing that the average joe wouldn't know. This educational effort is the brainchild of Doug Lenat, a computer scientist whose friendly manner and roly-poly appearance belie a remarkable intensity. That energy has been crucial in driving a project so ambitious that most of Lenat's fellow researchers have maintained that it couldn't possibly meet its goal; to equip a computer with all the general knowledge of the average adult. Considering that until now no computer has ever absorbed the knowledge of even the average four-year-old, this is a daunting task. But Lenat insists that, as he rounds out the sixth year of his ten- year project, he's right on schedule. "Teaching a computer all the ordinary things human beings take for granted is not an easy job", he says. "We have to put in ten times as much information as I though we would; but we're doing it ten times as fast." Computers, for the most part, are fantastically stupid things. Oh, they seem smart enough; no human brain could ever hope to compute pi to a billion decimal places, or contemplate several million chess moves and their possible consequences. But take a computer beyond these narrow areas of genius -- ask it "What is a tree?" or "Who is taller, a mother or her infant son?" or "If you paint a chair blue, is it still the same chair?" -- and the same machine is completely lost. Computers are utterly ignorant of the trivial knowledge humans spend a lifetime accumulating. If electronic brains are ever to get beyond the glorified bookkeeping and computing tasks to which they're largely confined today, they need programs that include at least a portion of this unglamorous data base. One approach towards this end has involved the development of "expert systems". Unlike standard programs, which tell a computer precisely what to do, an expert system incorporates heuristic knowledge -- rules of thumb that come only through experience. For example, an auto-repair expert system might be loaded with a few hundred if-then rules such as, "If the engine won't turn over and the temperature is below 20 degrees, then check the battery strength." After checking to see which of the conditions have been met, the computer employs the correct solution. A top-notch expert system can consider thousands of rules and come up with the same results that a human expert would. But expert systems, according to Lenat, are "brittle". "They have only a veneer of intelligence", he says. "When they're faced with a novel or unexpected situation, they'll give the wrong answer, and be convicing about it." To be wrong less often, computers need to be equipped with "consensus knowledge" -- knowledge that most of us take for granted. So Lenat and his team are entering 100 million morsels of general information into a program named Cyc (as in en-cyc-lopedia). And that's just for starters. In a few years, Lenat claims, Cyc will understand English well enough to continue its education on its own by reading books, newspapers, and magazines and discussing them with human tutors. Ulitmately, he asserts, Cyc could serve as a kind of knowledge fountainhead for other computer programs, allowing them to tap into a generalized data base and apply what they learn to their own particular areas of expertise. "My hope", says Lenat, "is that by 1999 no one would think of buying a computer without Cyc, any more than anyone would now think about buying a computer without word processing." Lenat could easily pass for the quintessential computer nerd. "Uh- oh", he says, when a video-game-like shriek erupts from one of the computer terminals next to his desk. With the quickness of a parent answering a baby's cry, he pushes off on his black Reeboks, spins his chair around, and attacks the keyboard with the blurred-key-click speed that programmers acquire from being impatient to get instructions into the computer. "Okay, we're in business", he says to the machine. For someone who is on such intimate terms with computers, Lenat learned his craft in a rather informal way. Most of his college training came from summer programming jobs in computer firms. It wasn't until he was ready to graduate from the University of Pennsylvania in 1971 that he took his first computer-related course. At the time, Lenat was polishing off the requirements for two undergraduate degrees -- one in mathematics and one in physics -- along with a master's degree in applied mathematics. Then he stumbled upon a course in artificial intelligence. "I had already applied to graduate schools in math and physics", he reccounts. "But AI was so fascinating that I decided that's what I wanted to do." Lenat made good in his new field at Stanford's celebrated computer science department. One of his earliest successes was a program called AM, a "machine learning" program designed to derive new ideas from a preprogrammed data base. Machine-learning programs are essentially computer contemplators: given a set of principles about, say, math or chemistry or baseball, the program turns the data this way and that to see whether it can come up with any new rules. We use this kind of reasoning every day: the chemist who invents a new polymer or the mechanic who designs a new engine is not so much creating new knowledge as manipulating what's already there. Lenat armed AM with such basic mathematical information as how to determine whether two sets are equal and how to perform the inverse of a mathematical function. When he turned the computer loose, it proceeded to make 700 "discoveries". AM had noticed, for example, that there was a certain subset of numbers that have unusually few divisors. Then it found that when it multiplied these numbers together in different combinations, it could find any whole number. Thus the program stumbled upon prime numbers and the unique factorization theorem. AM even uncovered a couple of mathematical principles that appeared to be genuinely original, including one that describes the relationship between the size of a prime number and the number of times is appears as a factor of certain larger numbers. (As it turns out, a human mathematician had already made the discovery, but at the time Lenat -- and his program -- did not know this.) AM earned the 25-year-old Lenat a prestigious AI award and sudden widespread recognition in the field. After receiving his doctorate and picking up faculty positions at Carnegie-Mellon and Stanford, Lenat went on to develop other machine- learning programs. But even the best of them kept running out of steam after an initial burst of discoveries. "The learning you get out of these programs is really only what you preengineer into them", Lenat says. "It's like a spring unwinding; after a while the initial programming is exhausted and the computer needs to be given more data. In 1983 I got this vision about what was holding back the program; it lacked common sense." In 1984 Lenat received an offer he couldn't refuse, as he puts it, to attempt to put his new theory into practice. The offer came from the Microelectronics and Computer Technology Corporation (MCC) in Austin, a research and development consortium funded jointly by Digital Equipment Corporation, Kodak, Apple Computer, NCR, and Bellcore. Lenat moved to Texas and immediately set to work to make the most out of the ten years and the $35 million MCC provided him. Despite MCC's conservative corporate sponsorship, Lenat has managed to set up a working environment that might have been lifted from any graduate school AI department. Young people in T-shirts and sneakers lean against doorways with labels like HACKER'S HELL and catch each other up on late-night computing feats. At lunch, which is shared by the group every day, arguments about obscure points of theory are spirited and sometimes punctuated by flying napkins and lime wedges. The air of informality aside, the task that confronts Lenat and his team of 12 is a rigorous one. Gathering knowledge isn't the problem. What's hard is finding a way to represent the knowledge so that a machine can make use of it. After all, Cyc can't learn about automobiles by going for a ride in the countryside. Each piece of knowledge must be carefully spoon-fed in exactly the right form. Lenat calls it teaching by brain surgery. The surgery is carried out in a sort of programming language developed by Lenat and his collaborator R. V. Guha specifically for its ability to handle the fuzzy concepts and irregularities that characterize the real world. Called CycL (and spoken by CycLists), it is based in part on the concept of "frames". A frame is a collection of facts and rules that tells Cyc what it needs to know about a "unit", or a particular thing or concept. The frame describing the unit Doug Lenat, for example, tells Cyc that he is an example of the unit Person, that he is a professor, that he belongs to the Democratic party, and that among the people he likes is Cyc administrative director Mary Shepherd (in fact, he lives with her, but apparently no one wanted to get into this with Cyc). On the computer screen a frame looks like a fact sheet; to build a frame about a particular unit, a Cyc "knowledge enterer" must first decide which kinds of information about the unit will be useful to the program. These information categories, or "slots", are then typed in as a list. The slots are then filled with "values". The frame for the unit Eating, for example, might look like this. EATING WHAT HAPPENS WHEN DELAYED: HUNGER PERFORMED BY: ANIMALS THINGS CONSUMED: FOOD MORE GENERAL PROCESS: CONSUMING The knowledge enterer must ensure that Cyc knows what all the slots and values mean. Thus Cyc needs to have a seperate frame constructed for Things Consumed and Food (if those don't already exist) and for every slot and value that appears in the frame. The units that are described by the frames can often inherit traits from other units. Thus, because the unit Chewing is an example of the unit Bodily Function, Cyc, automatically knows that chewing is one of the things that a person or animal is likely to do on a regular basis. Similarly, if Cyc knows that Bob and Jim are neighbors and that Bob's house is in Texas, it will know that Jim's house is probably there also. Cyc's ability to process several frames at one time helps it draw other relationships between units. If one of Cyc's frames tells it that World War II took place between 1939 and 1945 and another tells it that a particular battle between two of the war's combatants occurred in 1943, it will correctly conclude that the smaller clash was part of the larger war. A portion of Cyc's programming is set aside for processing logical constructions. If told that "Bill is either a terrific fisherman or a terrific liar", Cyc understands this to mean that the unit Bill has one or the other of these attributes, but not both. Cyc then consults what it knows about fishing, lying, and perhaps even egos and understands that fishermen like Bill may sometimes be motivated to fib about what they've caught. Like most AI programs that have to deal with the real world, Cyc uses default logic, which means that it can be told to assume something is usually true but that there may be exceptions. Thus, rather than having to specify to Cyc that horses, mice, people, and elephants give birth to live young, a programmer can simply inform it that almost all mammals give birth to live young. If Cyc then runs across a references to a platypus and learns this animal too is a mammal, it will guess that the platypus follows the birth rule; someone would have to tell Cyc that the platypus lays eggs. That may seem inexact, but, in fact, people reason this way, too. "If I asked you where your car was, you would assume that it was where you left it", says Lenat. "Your car may have just been stolen, but it's better to get an occasional surprise than to exhibit the irrational behavior that would result from always thinking that your car was stolen." Like we do, Cyc also has to know how to ignore irrelevant information, lest it waste too much time thinking or providing answers that are correct but useless. "When we're deciding whether to lock our car, we don't ask ourselves how many humps a camel has or what we had for dinner", says Lenat. "And if someone asked if there had recently been a large quantity of blood in your car, you would know not to take into account the fact that your body is a big bag of blood." Cyc has ways of avoiding such impracticalities. It knows, for example, to start its search with the most obviously relevant frames, and then to branch out to those frames that have the most in common with the original frames. Thus, if asked a question about freeways, Cyc might start with the frame for freeway, then move on to roads, cars, and so on, coming across its answer long before it got to railroads and camels. In the case of the blood question, Cyc would take advantage of its knowledge that in many situations people overlook the fact that they contain blood, allowing the computer to put that fact near the end of its list. To improve its speed further, Cyc is given access to about two dozen different types of reasoning, each of which is best suited for a certain type of problem. For complex problems, Cyc may sift through thousands of rules to reach an answer. But if asked how grandparents are related to their grandchildren, Cyc may simply apply inverse reasoning to its knowledge that grandchildren are their grandparents' children's children and conclude that grandparents are their grandchildrens' parents' parents. When information is incomplete or uncertain, Cyc can try to fill in the blanks and assess plausibility to come up with a best guess. If told that a person has voted and that the person's mother is 40 years old, for example, Cyc might guess that the voter is about 19 or 20. You've got to be at least 18 to vote, Cyc knows, but if you're much more than 20 you're not likely to have a 40-year-old patient. Bringing Cyc to this level of sophistication was an extraordinarily painstaking process. For the first year of the project Lenat and his team hardly went near a keyboard. Why bother trying to teach Cyc what a truck is if it doesn't understand cars, wheels, or even motion? The group had to start by defining such fundamental concepts as space, time, action, and physical objects. "We sat around for four months with enormous pieces of white paper", recalls Lenat, "trying to determine what categories were worth distinguishing. A lot of our insights were surprising, like when we realized that what we think of as tangible objects, such as a person or a table, are also events with a starting time, an ending time, and a duration. We don't like to think of it, but people and things are executing the transitory process of existence." The team also found it necessary to teach Cyc the complicated notion that people and objects have physical and nonphysical components that must be regarded seperately. If asked to provide the age of a book, for example, Cyc would have to determine whether it should consider the date that the physical book itself rolled off the presses or the date that the contents of the book came into being -- that is, when the author finished the manuscript. Many other abstract representations were programmed into Cyc in those first years. Time was explained in terms of 50 different types of relationships between events, such as, "Event A starts after the end of event B." Physical objects were broken down into things that are countable, like marbles, and things that are "stuff", like peanut butter. Continued in issue #200! ****************************************************************************** To join in the fun, send your name, home address, home and work phone numbers, and country preferences to Eric_S_Klien@cup.portal.com. ****************************************************************************** Up