o s 5i Ia Lynn Conway stands behind her workat XeroxPalo AltoResearch Center. Alan Bell (center) and Gaetano Borriello areplaying a game called "Truckin"' which tests artificial intelligenceapplications. "Truckin"' was designed by researchers working with Conway—Mark Stefik and DannyBobrow.

Creating a new generation of Palo Alto's Lynn Conway has been entrusted with mobilizing American computer scientists to build a new generation of machines that can think. She's out to explore the New Frontier of Supercomputing — before the Japanese stake territorial claims.

Michael Doyle

Conwaydescribesherselfas afan Research Projects Agency (DARPA). excitement about her work PA, Conwaywill scout the frontier ofsuper- Lynnof the "laid-back Palo Alto "Let's gomake it happen," Conway said in computer research and develop- computing. Othjer countries will provide lifestyle." Sitting in her well-tended in a tone reminiscent of an inspirational Conway'sment is as tangible in her writings as competition, particularly Japan, which has backyard with her Siamese cats Samantha coachbefore the big game. Conway willlead in her speech. In a lecture givenat Caltech embarkedon its own, highlypublicized ad- and Rapunzel, the 42-year-old computer a project with major defense and commer- and later published by Xerox,Conway said vanced project, scientist talksof sailing, bike rides and hik- rial implications. that the development of computers and The DARPA computer project for which ing in FoothillsPark. Working at the Defense Department, computer networks "arereminiscent of the Conway holds primaryresponsibility is en- As of this week, however,Lynn Conway Conway directs the government's effort to pervasive effects of the telegraph and the titled "Strategic Computing and Sur- has abandonedthe laid-back life for a more developa very very powerful computer railroads, as they spread out everywhere vivability." A total of million has been frenzied existence in Washington, that can reason like a human being and during the 19th century, providing an in- earmarked for the project in 1983-84, and where she heads a new national effort to understand everydaylanguages, frastructure people could use to go on another millionis scheduled for '84-85. t- developan advanced supercomputer. After 10 years of research work at the adventures, to go exploring and to send The supercomputers and associated Conway measures her phrases carefully Xerox Palo Alto Research Conway back news of what they had found. technologies developed through DARPA and listensintently,in themanner ofsomeone will not spend much time in the lab on her "I think of personal computers and the will serve military purposes, according to accustomedtoleadingothers.But she getsen- new job. Rather, she will be a pollinator, computer communication networks as a the DARPA R&D Program Summary thusiastic when discussing her favorite con- spreading money, advice and encourage- similar sort of infrastructure ... as we ex- Description, includingaircraft carrier com- cept—pushingforward the state of theart in ment among the various industry and plore the modern frontier the frontier of mand and control, photo interpretation, computer technology. "That'sreallycool!", universityresearchers working ongrants she what we can create." — pilots' assistant and strategic targetplanning. Conway is likely to say when somethingex- will administer, As computer project manager with DAR- But Conway is more excited about the cites her or catches her fancy. — chance to explore state of the art "I like to be part of the community of technology.She also wants to developthriv- people who go for it," Conway said, ing networks of advanced researchers the repeating the phrase "go for it" several 'There are some incredible technical social and technical infrastructure—that times in the conversation and sometimes forms the basis for future exploration. The pounding her fist into her palm for em- opportunities that can be taken if we go for actual supercomputeris but one goal of the phasis. it. We don't have to form some kind of project. In an exclusive interviewwith the Weekly "Itgreatly oversimplifiesto saywe'reo\it before she leftfor Washington, Conway ex- institute; wherever the researchers are, they to produceamachine,"Conway said. "Any pressedthe same can-do attitudetoward the can participate." one machine is only onepoint in the design challenges ahead of her as computer space." researchmanager for the Defense Advanced Conway observedthat, in the traditionof Palo Alto Weekly " August 3, 1983 " Page 11

fast, $50 DC, $95

Center,

______^_ - Supercomputers Tang and Space Blankets, theresearch con- ford. Feigenbaumrefers to the MCC pro- ductedfor thegovernmenton the suDercom- posal as "an American Fifth Generation." puter will lead to diverse and unexpected 'The charter of DARPA is, in a sense, to At the Center for Integrated commercialdevelopments. explore the farthest Systems (CIS) is a consortium of 18 elec- "You'llsee a wholearray of technologies frontiers ... to have a tronics firms (including Xerox and Hewlett and knowledge"spinoff from the DARPA sense of what is possible." Packard), which put up $750,000 apieceto work, Conway predicted. Fueling these conduct research. DARPA has provided a spinoffs will be the "col- three-year, million contract to CIS. The laborative/competitive" R&D environment annual CIS research budget is about that Conway hopes to establishthrough her million—a net 13.5percent increase over the apart is thecomputer's ability toreason and million. work at DARPA. This environment com- 1983budget. Currently, major projects are to understand everyday language. Such Conway noted that almostall ofthe work bines new technologies like computer net- working on advanced cruise missiles, laser computers can be providedwith alargebase sponsoredby DARPAis actuallycontracted works, with the desire to create and the weapons and what the summarydescription of knowledge. They can make intelligent out—to industry,to universitiesand to corn- desireto be number one. reports as "a major new initiative in a new decisions, bined groups like CIS. generationofsuper speed computers." This Some so-calledexpertsystems are already Conway's old job at the Xerox Palo is Conway'sproject. in practice. The Palo Alto-based firm onway speaks of attracting a Alto Research Center (PARC), located Stanford professor Edward Feigenbaum Teknowledge, founded by Feigenbaum, B "critical mass" ofresearchers in the Inin the hills of Stanford Industrial Park, and writer Pamela McCorduck, in their developedan expert system called Drilling different "compartments of she was involved with research in informa- book The Fifth Generation, explain that Advisor. This system diagnoses problems knowledge," such as software, chip design tion processing and related technologies. computer generations aredetermined by the encountered by oil well drillers and makes and naturallanguagecapabilities.Shehopes PARC has areputation as a well-funded, computer's central technology, recommendations for correcting the prob- therewillbe enough"interface" among the highly advanced research station, where the The first generation of computers was lem. groups that their workwill feed the totalef- scientists and engineers are not necessarily based on vacuum tubes. The second genera- This is the direction toward which the fort. obligated to produce immediate, commer- tion was based on transistors, and the third DARPA "strategic computing and sur- Conway specifically wants to stimulate cially viableresults. generationon integrated circuits. vivability"project is heading! The Japanese "an open evolution of knowledge." The "DARPA is to the Department of Integrated circuits use many transistors areheadingthat waytoo. Much ofTheFifth best encouragementfor R&D advancement, Defense as PARC is to Xerox," Conway Packed ontoa singlesilicon chip. The fourth Generationdescribesa 10-year, million she says, is for competing designs to meet said. Like "the charter of DARPA generation of computers, which Feigen- Japaneseproject to developafifth genera- openly, is to explore the, farthest frontiers" in baum says will "dominate during the tion computer. Designs thatworkcan be imitatedand im- various scientific disciplines. According to 1980s," is based on Very Large Scale In- "It's not just a response to the plemented morerapidly than if theresearch the DARPA summary description, this is tegrated (VLSI) chips. Japanese," Conway said of the DARPA were conducted in isolated, secretive in- ' 'high-risk research and developmentwhere VLSI is atechniqueforcramming 100,000 computer project. More than one team of dustry think tanks. Researchers can share major improvements in militaryoperations or more integrated circuits onto a single is already working on so-called ideas even though they're working in dif- could evolve." chip. Much of Conway'sreputation isbased fifth generation projects. ferent parts of the country by using tools DARPA was founded in 1958, in large on her work in VLSI design, Ten American corporations (Digital like the computer network. Computer net- part aresponse to the Soviets launchingthe The supercomputer that Conway will at- Equipment and Control Data among them) works are recent innovations, "a strong Sputnik satellite.DARPA playeda "signifi- tempt to developwillpresumably represent formed the Microelectronics and Computer tool," inConway's phrase , thatcan quicken cant part," according to the summary the fifth generation. It will be based on Technology Corporation (MCC) in 1982. other innovations. Conway built her con- description, in developingsuch items as the VLSI chips and willoperate veryfast- which recently established its head- siderable reputationthrough theuse ofcom- Stealth bomber, strategic cruise missiles, handling more than 100millioninstructions quarters in Austin, Texas, will undertake puter networks. ABM systems and early space satellites. per second. research in computer design that would be In the '70s, Conway was attempting to DARPA's 1984 budget request is But what really sets the fifth generation too expensive for a single company to af- design a new type of facsimile transmitter;

SUMMIT * SPORTS CAMPS Dy the Week TEN SESSIONS JUNE 20 - AUGUST 26 BUS SERVICE FROM ALL SCHOOLS TENNIS LESSONS THREE HOURS/WEEK HALF-DAY CAMP 12:30 - 4:30 BOYS & GIRLS Ages 7-11 at Jordan Ages 11-15 at Paly All Ages " Cubberley

Mondays through Fridays 9:30 am -4:15 pm Sign Up By-The-Week at Jordan Middle School Every Monday at 9:00 am - Tennis Courts - $77 per Week Athletes Grouped by Age and Weight

Telephones: 328-2299 and SOCCER.BASKETBALL.TENNIS.SWIMMING LACBOSSE.BASEBALL.TBACK.DODGEBALL

Page 12 " August 3, 1983 " Palo Alto Weekly

*

Stanford,

$8 $12

$450 PARC,

MCC,

$867.7 "

an instrument that can transmit number of silicon chips involved. That set tookplaceoverthe ARPAnet, the computer Massachusetts Institute of Technology, photographsfrom one locationto another. Conway to thinking about VLSI chips. network established by DARPA that con- Students used the ARPAnet to transmit "I came up with a very elegant architec- Conway and from the nects industries and universities affiliated their designprojects to Xerox which .tural solution," Conway recalled. Unfor- California Institute of Technology began with DARPA. arranged for the chips to be built. Because tunately, the solution would also have been collaborating on simpler ways to design In 1979, Conway used ARPAnet toteach of the computer network,the students were prohibitively expensive because of the VLSI chips. Much of their collaboration course in VLSI design at the able to see a finished product much quicker Calling for a national effort University computer scien- a "leadership challenge something like chooses his wordscarefully, patientlyex- tist Edward Feigenbaum has what it tookto build the Stanford Linear plaining basic concepts of computer Stanfordpopularized the notion that the Accelerator. science to head off misinterpretations. United States needs a coordinated na- "It takes a national will, or at least a Feigenbaum the entrepreneur has tionalstrategy to developthe supercom- regionalwill," he said. "It takes a vision helped found two Palo Alto companies puter, the project Lynn Conway is im- of whereit has to go, and a tremendous that apply artificial intelligence work. plementing. amount of energy to pull it together." Teknowledge designs problem-solving Feigenbaum's The Fifth Generation Feigenbaum sees no heroes on the computer software packages for clients (written with Pamela McCorduck), drew scene, but described what that leader like the French national oil company, national attention to the field of ar- would have: "A single-mindedness of which paid millions for a program to tificial intelligence research when the purpose, scientific breadth, respect in spot problems with oil wellsand to sug- book was issued this spring, only weeks the community and charisma." gest ways to correct and avoid them. before Conway's appointment. Feigenbaum counts himself out as His other The book focuses on coordinated ef- that leader, preferring to split his time designs computer programs for the forts by Japanese industry and govern- among teaching, writing technicalbooks emerging geneticengineeringindustry. ment to overcome the current two-year and popular ones like The Fifth Genera- Those firms arebut two of many Palo lead that the United States holds in ar- tion, research in artificial intelligence Alto area companies involvedin the ar- tificialintelligence research. Feigenbaum and entrepreneurial efforts in the field. tificial intelligence field. Feigenbaum warns of the devastating consequences He takes to his roie as popularizer believes this areacould emerge asa major to the U.S. economyif the Japanesesuc- gleefully, but has discoveredthe difficul- center for artificial intelligence activity. ceed in developing the new technology, ty of explaining abstruse concepts to "There's no coordinated but creating a world market for it and then non-technical audiences. On the radio- since we all live in PaloAlto, we see each dominating that market. television talk show circuit (which in- other a lot," he said. "We are writing this book because we cluded an appearanceon the Merv Grif- Those uncoordinated Feigen- are worried," Feigenbaum states in his fin Show), he has encountered questions baum believes, must be drawn together preface. like, "Why should we care about what in a more systematic way—and soon. The coordinated effort needed in the you do?"In simplifying his complicated United he said in an interview, is specialty for lay audiences, Feigenbaum Stanford's EdwardFeigenbaum —by Julie Jacobs

PARC,

firm, IntelliGenetics,

effort,

efforts,

States, *

5 Supercomputers than they would have otherwise. are the notions ofsimplifying technicalpro- Conway traced her interest in the social, entered the workplace. Distributing infor- That quick turnaround time is part of cedures and making them accessible, and networking aspects of computer research mation to stimulate further research was what Conway referred to as "the open utilizing powerful tools like computer net- back to her undergraduate years at Colum- then accomplished through copy machines, evolution of techology." With products in works. bia University, where she tooka number of as it was laterthrough computer networks. hand, researchers can more quickly test In the last several years, Conway has anthropology courses. She earned a B.S. Conway's term for this is "intellectual which designs work—and Mead and Con- leverage." way could modify their chip design Through funding, her personal contacts, methods. conferences and the strong toolof the com- The Mead-conwayMead-Conwaymethod for designing "I'm trying to push the state of the art, to puter networks, Conway now aims to cause the new things to happen. (I like to) be leverage advances in the field of advanced ffi?lS^-3!S'*hi!*£ . supercomputers. rather than depend on specialists. The part of the community of people who go for it." ''There are some incredible technicalop- method is currently taught in about 100 -'■■'-. portunities that can be taken if we go for universities. it," Conway enthused. But her own "In my opinion," said John Linville,co- worked in the field of artificialintelligence, and M.S. in electricalengineeringfrom Col- challengeswill not strictly be technical. She director of the Center for Integrated She is attemptingto simplify therules of ar- umbia, then workedfor IBM and Memorex will represent the DARPA supercomputer Systems, "in the last five years, (Conway's) tificial intelligence and make the field "ac- before joining Xerox in 1973. project to policy makers and budget book has made more impact in the area of cessible to the averagecomputer scientist." Tool use another concern of an- granters, like Congress. integrated circuits than any other develop- Some of that work has been with the thropologists—has also intriguedConway. "That'sanothersphere ofknowledgeand ment." Heuristics Programming Project at Stan- She recalled "being one of the early copy experience" to explore, she said. For Con- But equally important, Conway stressed, ford freaks" when the first copy machines way, it's a new frontier.®

f 1 — 1 .iaffure. Announces the addition to our staffof: MR. MARCUS (Vidal Sasoon's Hair Cutting Method) Formerly of Vidal Sasoon Salon in Beverly Hills, California, he is a specialist in new hair coloring technique for a personalized, healthyand attrac- tive look.