conduitcondu t - Brown University · turn (greedy), whereas others made glo-bal decisions regarding all their clients ... Next year’s competi-tion will be held in conjunction with

Volume 9, Number 2 Department of Computer Science Fall, 2000 Brown University

Brown University, Box 1910, Providence, RI 02912, USA

conduit!condu t!$295”; in these auctions, the (NP-complete)problem of determining how to allocate thegoods so as to maximize revenue falls tothe hands of the auctioneer. In simulta-neous auctions, however, the complexityburden lies with the bidders.

The first international trading agent com-petition (TAC), organized by the AI labora-tory at the University of Michigan and led

by Mike Wellman, challenged its entrantsto best-design a trading agent capable ofbidding in simultaneous on-line auctionsfor substitutable and complementarygoods. TAC was a success, as 22 agentsfrom around the world entered the compe-tition; 12 of them qualified to compete inthe finals, which were held at ICMAS ’00 inBoston this past July. My agent, RoxyBot,developed in collaboration with JustinBoyan of NASA Ames Research, finished ina four-way statistical tie for first place.

RulesThe TAC competition consisted of a seriesof game instances, each of which pittedeight autonomous bidding agents againstone another. Each TAC agent simulated atravel agent with eight clients who were

Suppose you want to buy aused Canon AE-1 SLR cam-era and flash at an on-lineauction. At last count, over4000 links to on-lineauction sites were avail-able at advocacy-net.com. It would be adaunting task to manu-ally monitor prices andmake bidding decisionsat all sites currentlyoffering the camera—especially if the flash ac-cessory is sometimesbundled with the cam-era, and sometimes auction-ed separately. But for thenext generation of automatedtrading agents—bots—this

will be a routine task.

Simultaneous auctions, which arise natu-rally on the Internet, are a challenge tobidders, particularly when complemen-tary and substitutable goods are on offer.Complementary goods are items such as aflash, a tripod, and a case, that wouldcomplement a camera—but a bidderdesires any of the former only if s/he iscertain to acquire the latter. Substitutablegoods are goods such as the Canon AE-1and the Canon A-1—a bidder desires oneor the other, but not both. In combinato-rial auctions, bidders bid on combinationsof items, such as “camera and flash for

INTERNET AGENT ECONOMICS:A Trading Agent Competition

Amy Greenwald

“Simultaneous auctions,which arise naturally on theInternet, are a challenge to

bidders...but for the nextgeneration of bots they will

be a routine task”

conduit! 2

ment tickets on a given night are substi-tutable (a single client can go to at mostone of symphony, theater, and baseball);thus, an agent had to trade off among itsclients’ preferences for the various typesof entertainment tickets and their pricedifferential. Similarly, the good and badhotel rooms were substitutable: an agenthad to trade off between its clients’ pref-erences for the good hotel and the pricedifferential between the two hotels.

StrategiesThe most substantial strategic dichotomyin the competing TAC-agent designs wasin the use of greedy vs. (approximately)optimal decision-making algorithms.Greedy algorithms are simple to imple-ment and fast, but in general exhibit sub-optimal performance. Optimal algorithmsmust repeatedly solve NP-complete prob-lems in real time. For example, whilesome agents focused on obtaining com-plete packages, others made bidding deci-sions on travel packages alone (i.e., flightsand hotel rooms) without regard forentertainment tickets, essentially break-ing the TAC problem down into two sub-problems, and solving greedily. Althoughsimpler, the greedy strategy is not opti-mal in general. In particular, it is pref-erable to extend a client’s stay wheneverthe utility obtained by assigning that cli-ent an additional entertainment ticketexceeds the cost of the ticket and an addi-tional hotel room (plus any travel penal-ties incurred). Similarly, it is sometimespreferable to sell entertainment ticketsand shorten a client’s stay accordingly. Asecond example of this strategic dichot-omy was seen in the fact that some agentsaimed to satisfy each of their clients inturn (greedy), whereas others made glo-bal decisions regarding all their clients’interests simultaneously (optimal).

RoxyBotRoxyBot is short for “ApproximateBot,” aname that represents our goal of con-structing a trading agent whose strate-gies approximate optimal behavior. UsingAI heuristic search techniques, RoxyBotincorporated an optimal solver for theproblem of allocation—assigning alreadypurchased goods to clients at the end ofthe game so as to maximize total utility.RoxyBot also incorporated an optimal

interested in traveling from TACtown toBoston and home again during a five-dayperiod. Each client was characterized by arandom set of preferences for the possiblearrival and departure dates, hotel rooms(The Grand Hotel and Le Fleabag Inn),and entertainment tickets (symphony,theater, and baseball). A TAC agent’sscore in a game instance was the differ-ence between the total utility it obtainedfor its clients (i.e., measure of success asdetermined by its clients’ preferences)and the agent’s expenditures.

In order to obtain utility for a client, anagent had to construct a complete travelpackage for that client by purchasing air-line tickets to and from TACtown andsecuring hotel reservations. It was alsopossible to obtain additional utility bysupplementing a travel package withentertainment tickets. Each item wassold separately at auction: in total therewere 28 goods, and therefore 28 simulta-neous auctions. Airline tickets were soldin single-seller auctions that cleared con-tinuously. Hotel room reservations weresold in simultaneous ascending ‘English’auctions (auctions like those for antiquesand art). These auctions could clear atrandom after a sufficient period of inac-tivity, but otherwise cleared at the end ofthe game. Entertainment tickets weretraded in continuous ‘double’ auctions(auctions like those on the New YorkStock Exchange).

Complete details are available at:http://tac.eecs.umich.edu.

DecisionsDuring a TAC game, agents were continu-ously faced with three basic decisions:what goods to bid on, how many to bid for,and what price to bid at. These biddingdecisions were complicated by the trade-offs resulting from complementary andsubstitutable goods. For example, a flightto Boston complements a flight back toTACtown, but a one-way ticket is of noutility. Similarly, an entertainment ticketfor a given night would complement atravel package that included that night, aflight arriving that day, a hotel room forthat night, and a flight departing the nextday—but the entertainment ticket is ofvalue only if the complete travel packageis obtained. Moreover, multiple entertain-

conduit! 3

-2000

-1000

0

1000

2000

3000

4000

5000

6000

ATTac RoxyBot Aster UMBC ALTA DAIhard RiskPro T1

Sco

re

Final Round (13 Games)

-2000

-1000

0

1000

2000

3000

4000

5000

6000

RoxyBot Aster DAIhardATTac RiskPro UMBC ALTA Gekko T1

Sco

re

Preliminary Round (~70 Games)

trage in a unified way. RoxyBot’s high-level strategy is outlined in Table 1.

ResultsThe results of the TAC competition aredepicted in the graphs below: the firstgraph depicts the scores in qualifyingrounds (up to 90 games, with the lowest10 scores dropped), and the second graphdepicts the scores on competition day (13games). Along with RoxyBot, the otherthree top-scoring teams were ATTac,Aster, and UMBCTac. ATTac, built by ateam of researchers at AT&T (includingMichael Littman, Brown PhD ’96), is anagent whose functionality is best charac-terized as adaptable; its flexibility en-abled it to cope with a wide variety of sce-narios during the competition. Aster,developed by Brown CS’s IPP Partnerintertrust.com, is an agent that is nei-ther strictly greedy nor strictly optimal;scalability, rather than optimality, wasforemost among its designers’ goals, sincethey expect many situations of practicalinterest to be more complex and lessstructured than TAC. UMBCTac’s competi-tive edge is that it conserves networkbandwidth; on average, this agent up-dates its bidding data every 4-6 seconds,providing a significant advantage overthe reported 8-20-second delays experi-enced by competing agents. There was a

solver for the more general problem ofcompletion—determining the optimalquantity of each item to buy and sellgiven current holdings and forecastprices. The formulation of the completionproblem involves a novel data structurecalled a priceline designed to handle (esti-mated) prices, (estimated) supply anddemand, sunk costs, hedging, and arbi-

Mean, minimum and maximum at 95% confidence interval

Table I

(A) While enough time remains for atleast 1 bidding cycle, do:1. Update current prices and holdings

for each open auction.2. Based on current and historical price

trajectories, estimate closing prices andthe supply and demand of each good;store these estimates and current hold-ings as a priceline.3. Run the completer to determine theoptimal quantity of each flight, hotel,and entertainment ticket desired; com-pute the difference between the optimalvalues and current holdings.4. Set bid/ask price strategically to

buy desired goods at minimum cost andsell undesired goods at maximum prices.

(B) After all auctions have closed, runthe allocator to assign goods optimallyto individual clients.

conduit! 4

rather than routine applications of exist-ing technology.’ ‘This initiative,’ said Pres-ident Clinton in announcing the awards,‘will help strengthen America’s leader-ship in a sector that has accounted forone-third of U.S. economic growth inrecent years.’ Brown Computer Scienceinvestigators received two of the only 62‘large’ (more than $500,000) ITR grantsawarded nationwide.

Prof. David Laidlaw’s four-year, $2.3-mil-lion grant will fund an interdisciplinaryand inter-university research team thatincludes Brown’s Andy van Dam, an art-ist, an applied mathematician, a biomedi-cal engineer, biologists, and perceptualpsychologists who will develop new waysto look at scientific data from magneticresonance imaging (MRI), computer-sim-

TAC panel at EC ’00 in Minneapolis thisOctober, in which the top four TAC agentteams participated. Next year’s competi-tion will be held in conjunction with EC’01.

Artificial Intelligence andEconomicsMy graduate seminar this semester,Internet Agent Economics, is concernedwith the use of game theory and econom-ics as frameworks in which to model theinteractions of Internet agents. It covers

both the design of Internet agents and theanalysis of the potential impact of billionsof such agents on technology and society.Selected topics include web auctions,automated negotiation, recommendersystems, and shopbots and pricebots.With regard to the pedagogical aspects ofTAC, graduate students may satisfy theirprogramming requirement by implement-ing a (approximately optimal) TAC agent,while the final project in CS 141, myspring-semester undergraduate AI class,will be to build a (greedy) TAC agent.

In mid-SeptemberNSF announced theaward of its presti-gious InformationTechnology Research(ITR) grants.

The major goals of theprogram are to aug-ment the nation’sinformation technol-ogy knowledge baseand strengthen its ITworkforce. NSF dir-ector Rita Colwell

said, ‘These projects represent majorinnovations in information technology,

“THE UNIQUENESS OF CS92” borne out ...In the spring ’99 issue of conduit!, Roger Blumberg wrote about CS92, the Educational Software Seminar, in whichour undergraduates create instructional software for local schools including Vartan Gregorian Fox Point ElementarySchool. One weekend afternoon in September, he sent the following letter to the four students who worked on one of thespring 2000 projects for the Gregorian School:

Dear Roberto, Ranyee, Kevin and Imeh,

A few hours ago my doorbell rang and it was one of the kids in the neighborhood selling giant candy barsto support school field trips. It turned out the school was Vartan Gregorian and he was in the thirdgrade, so I asked who his teacher was this year and sure enough it was Karen, so I asked him what hewas learning in math (I didn’t say anything about computers). Well, the kid’s face immediately lit up andhe launched into a giant, breathless sentence about your Mad Math Minute, about how they use it atschool and he even got to bring a copy home and put it on their PC and now he uses it at home, and onand on. His father was with him as the kid walked around pushing the candy bars, and added “Yeah, it’sreally a pretty cool program, and you know some Brown students made it!” I said yeah, I knew. I wantedto tell you all about this, since it really is a testament to your having created something that not onlyworks but genuinely gets the kids excited. The kid’s enthusiasm and comments really made my day(more than the 4-pack of Reese’s!), and I hope you’ll take some pride and pleasure in it too. I also hopeyou’re all well and good and having an excellent semester. All the best,

Roger

CS AWARDED TWO NSF ITR GRANTS

David Laidlaw

conduit! 5

The original creators of the Artemis Project were AnneSpalter and Andy van Dam, as part of their NSF STCoutreach program.Encouraging women to pursue a high-tech career means fostering interest incomputers and technology duringtheir early teens. To this end, CSchairman Tom Dean, organizer of thefour-year-old Artemis Project,recruited four Brown students fromthe department’s Women in ComputerScience (WICS) group to run a five-week summer camp for girls finishing

the 8th grade who were interested in sci-ence and technology.

“If you want to encourage young womento get excited about the field,” says TomDean, “it’s probably not a good idea to geta bunch of stodgy old computer scienceprofessors together.” Instead, he called

upon Christine Davis ’02, Seema Ram-chandani ’02, Maryam Saleh ’01 and Glo-ria Satgunam ’03 to coordinate the projectand mentor the girls. To launch theproject they got a $15,000 donation fromthe Microsoft Local Community Alliance,$7,500 from the Brown University Lead-ership Alliance, and further support fromBrown’s Undergraduate Teaching andResearch Assistantship program. Namedafter the Greek goddess of the hunt,known for her strength, independenceand courage, the Artemis Project aimed tointroduce concrete computer skills andabstract computer science concepts whiledeveloping self-confidence and leadershipin an encouraging and challenging envi-ronment. Similar all-girl camps are be-coming a popular method to overcomewhat experts call the digital gender div-ide, the relative absence of women fromscience and technology-related fields.

ulated blood flow research, and geo-graphic remote sensing.

‘Current MRI technology, for example,can supply huge amounts of data, butthere is no practical way of visualizing it,’Laidlaw said. More information could beextracted from an MRI if one could some-how look at the multiple values repre-

sented at each point of the image.

The team will apply ideas frompainting, sculpture, drawing andgraphic design to create new visual-ization tools that portray hugeamounts of data as effectively as pos-sible. ‘We’re looking for more expres-sive pictures than are currentlyavailable,’ Laidlaw said. The re-searchers will use perceptual psy-chology to compare the effectivenessof visualization tools in several envi-ronments, including virtual reality,desktop workstations, paper and 3Drapid-prototyping output. The toolswill be developed and evaluated in

close collaboration with scientists in threedisciplines: neurobiologists studying neu-ral development and diseases via biologi-cal imaging, computational flowresearchers studying blood flow througharteries, and geographers using remote

sensing for environmental monitoringand natural resource management.

Under Prof. Stanley Zdonik’s five-year,$3.2-million grant, researchers (CS’sSteve Reiss, Michael Franklin of UC Ber-keley, and Mitch Cherniack, CS PhD ’99and now on the faculty at Brandeis) willstudy ways to make using the Internetfaster and more responsive by designingweb-based middleware—software to en-hance the interaction between users andWeb servers. The team’s research focuseson designing technology to let people cre-ate profiles of their information interestsand on developing techniques that canuse those profiles to manage web dataintelligently.

Another goal of the research is to explorehow user profiles might let people updatethe information they carry in their porta-ble computers and cell phones. The teamhopes to develop techniques that wouldallow someone to plug into an Ethernetsocket to update the limited memory of acomputing device. The user would receiveupdated e-mail and other information.The researchers hope to make such “datarecharging” on a portable device as sim-ple as recharging a battery.

THE ARTEMIS PROJECT

Stan Zdonik

Project organizer,CS chairman

Tom Dean

conduit! 6

Artemis coordinators, l to r: Gloria Satgunam ’03,Seema Ramchandani ’02, Maryam Saleh ’01,

Christine Davis ’02

To promote the program, the Brown stu-dents visited area schools and met withguidance counselors. They were wellreceived by instructors who saw that thecamp’s activities and environment wouldnurture the girls’ curiosity, teamwork,self-confidence and creativity. Said Brownstudent Seema Ramchandani, “We wantgirls to become confident in themselves.They are between ages 12 and 16. That’sthe age that most female teens are goingthrough a lot of difficulties.... There arestereotypes that doing computers is notcool.... We have to help them have the con-fidence to get up there and say, ‘I can do itand not put society’s values before myown.’”

Forsaking early-morning television andfun with friends, 17 teenage girls spentfive weeks in the department’s computerlabs, building Web pages, robots, design-ing computer programs and developing anenthusiasm for everything high-tech. Themain focus for the Brown students was toteach the girls that technology is anenabling tool; to encourage them to bedesigners of technology and dispel theperception that it is an antisocial and soli-tary activity. Each participant wasassigned an undergraduate female men-tor from the WICS group with whom theycorresponded by email. In addition tousing logic and deductive reasoning inexploring topics in CS such as object-ori-ented programming, AI, smart robots andcomputer animation, theylearned the practical skillsneeded to use word process-ing, email and the WorldWide Web. During camp, thegirls were divided intogroups of five with aninstructor. One group builttwo robots, programming oneto follow a black line whilethe other lifted its hat whenyou shook its hand or wavedin front of it! Another teamworked on computer gamesfeaturing pirate ships,another designed computermaze games, and yet anotherwaded through a huge pile ofLegos to create a candy dis-penser! To build teamworkand leadership skills, thestudents participated in

more traditional camp activities—negoti-ating a rope course, throwing water bal-loons and being blindfolded and ledaround by a partner. Despite challengesand initial aggravations, the girls beganopening important lines of communica-tion among themselves and with theirinstructors and soon came to appreciatethe benefits of cooperation and teamwork.

Talks by CS faculty and alumni as well asfield trips to the Boston Science Museumand the CS virtual reality lab roundedout the schedule. Talks on art and sci-ence, computer animation and women inscience and engineering stimulated muchactive discussion. A visit by Danah Boyd’99, now at MIT’s Media Lab, helpedspark lively conversations about womenand technology—Danah is an ardent fem-inist. Beside the hardcore focus on tech-nology, there was plenty of time for justplain fun—a breakdancing session (nowthey can all jenga and down rock!), aspontaneous pajama party at a ThayerSt. Indian eatery after the daily infusionof lunchtime sandwiches began to pall,and many whacky trips to the supermar-ket. The high point of the five weeks wasthe last day, when parents, friends andsiblings arrived for a big party (shrimp,not sandwiches at this event!) and thegirls were able to show off their projects,web pages and robots. Thanks to thegreat generosity of Microsoft, each girlwas given a huge amount of software totake home. The instructors had asked

conduit! 7

puters at such an intense level (both theArtemis girls and some of the CS148 stu-dents, since there are no prerequisitesfor the course) requires a different mind-set from the one I have when workingwith my peers. I will be able to use theexperience I gained with Artemis whenthinking about the best way to teachthese concepts to the CS148 students.

Gloria SatgunamThis was definitely a one-of-a-kind expe-rience for me. Artemis was so much funas well as so much work. I have alwaysenjoyed working with kids, but beingresponsible for maintaining the focus of17 young girls six hours a day proved tobe quite a challenge. However, theinnovative way in which we introducedthe girls to computers and roboticshelped a great deal. From Lego Mind-storm® robots to Klik and Play’s graphi-cally oriented programming languageto our own interactive games, I believethat we succeeded in providing aninteresting and fun environment inwhich the girls could learn and have funsimultaneously.

Maryam SalehMy goal for this program was to “cre-atively” engage a group of young girlsinto learning more about science. Toachieve this we designed an educa-tional environment with no competitionand no absolute authority. The students

Microsoft to donate software so they couldgive each girl a parting gift. A list ofchoices was submitted (please send eitherthis or this, etc.), and Microsoft sent theentire list, enough for 20 students!

For the four Brown students, the Artemisexperience required a major commitmentof time, effort and patience, but eachgained insight and not a small amount ofself-knowledge, as the following reactionsattest:

Christine DavisThe last three months have been crazy!From the work we put in last semesterrecruiting at various middle schools inand around Providence (with no directaccess to a car, mind you) to trying totie up all the loose ends and move for-ward full speed with CS148 (BuildingIntelligent Robots), things have been alittle hectic, to say the least. But throughall of our adventures we have managedto pull through, coming closer togetherand learning a lot along the way. We’vecome a long way in our ability to workas a team—I’ve also gained a lot ofrespect for my teachers and professors. Idon’t think I ever knew how hard theywork or how much energy they put injust to teach their often recalcitrant stu-dents. Teaching also requires morepatience than I ever thought possible.Working with students who have neverprogrammed before or dealt with com-

The coordinators designed this lively and informative poster as afun way to document the many facets of the five-week program

conduit! 8

weren’t awarded grades for complet-ing a project: we trusted them and theirself-motivation. Furthermore, to elimi-nate the impersonal aspect of a stu-dent-teacher relationship, we usuallysplit the class into groups of five accom-panied by an instructor.

Still, there were times when lack of inter-est called for more motivational speak-ing on our part. When the robots wouldfall apart, so would the girls’ drive. This iswhere our responsibility kicked in tokeep them interested. Whether it be bymaking clowns of ourselves or the use ofanother tactic, the important part wasto stay honest. Each instructor was sim-ply a more experienced group memberproviding the means necessary tocome up with a solution. I avoidedshowing them a simpler path to the fin-ish line. And yes, sometimes the girlscame up with innovative ways of solv-ing a problem, and sometimes not. Butsince they had experienced their wayand had seen it didn’t work, they weremore willing to hear another method. Inoticed in our discussions that most girlshad inquisitive minds that questionedwhat was offered to them as fact. Theyalways challenged us to have solidproof behind everything we taughtthem. Hence, they motivated us to be

well informed and not take anything forgranted. It was a reminder of what Iconsider one of the most important les-sons in life.

Seema RamchandaniCertain things you learn through class-rooms, others are learned throughexperience. During Artemis I had achance to experience a new nichewithin the community as a coordinator(which includes being a team member,mentor, teacher, student). After the firstfew days, we realized that the key tomaking a day successful was to assign agoal for the entire group and worktogether to realize it by the end of theday. In a group, each woman supports,checks up on and becomes a resourceto the others. Within the first week, itseemed like the girls had also learnedtheir first lessons on “working in groups”.On the fourth day, I had given them theassignment of organizing themselves bytheir birthday while blindfolded andwithout speaking. At first, seventeen girlsspent fifteen minutes on the green out-side our building complaining (breakingthe silence rule) that it was impossibleeven to try. After realizing it was impossi-ble to gain my sympathy, one girl imple-mented an idea of communication andit spread like wildfire; with encourage-ment from us, they completed the taskwithin 15 minutes. The girls might losecomplete hope in completing the mis-sion that you gave them, or you mighthave a limp hair day, fall flat on yourbutt during breakdancing, or dropthree robots on the floor simultaneouslyafter each group spent four hours build-ing them, but the great thing was thatyou could take a five-minute break andthere were three other women to helpyou out.

One of the Artemis teenagers is now tak-ing a well-rounded course load at SheaHigh School, including classes in busi-ness, math and science to prepare for ahigh-tech career. She hopes to attendBrown, as do most of the participants.Says chairman Tom Dean, “We’d beecstatic, if the class of 2008 had a fewmore women graduating in computer sci-ence as a result of this summer’s ArtemisProject.”www.cs.brown/orgs/artemis

Programmed to move and react to light sensors, this robot wasmade by the the team of Jamie Burr, Classical High School;

Amanda Boyanowski, Shea High School; Nancy Tavares, CranstonEast High School; and Nicole Terrien, Seekonk High School

conduit! 9

great neighbors, such as Hillary, althoughit gets a little annoying when her hus-band comes for a weekend and we have towait for them on the highways.

Nevertheless, I live just half an houraway from Grand Central Station so atrip to Broadway and a nice dinner arerelatively common. As you can imagine, Ihave not even started to explore every-thing that Manhattan has to offer, but Ihave already learned a few lessons. Not

JOSÉ CASTAÑOS, PhD ’00It feels like a long time since Commence-ment. The week after graduation I joinedIBM Watson, just north of New York City.This region is very nice, with large forestsfor hiking where you often see deer, andthe Hudson is not far away. We have some

In mid-October, Professor Emeritus Peter Wegner was awarded the Austrian Medal

of Honor for Science and the Arts, First Class. The ceremony, attended by senior col-

leagues, was held at the Faculty Club. Austrian Consul General Dr. Harald Miltner,

who bestowed the medal upon Peter, thanked him for his work in helping Austrian

computer scientists. CS chairman Tom Dean described Peter “as a god in the pro-

gramming languages pantheon”, one of the first to explore the connection between

programming language, which controls how computers function, and the hardware

of computer machinery. Said Dean, “His research is so vital, that just recently, two

talks during an academic computer conference in South Carolina were largely

devoted to Peter’s latest work.” After the award ceremony, Dr. Miltner and his wife

accompanied Peter and Judith for cake (shaped like Peter’s award) at a more casual

get-together in the department, attended by the rest of the faculty, staff and students.

The great wonder for us all is that Peter, who survived the Holocaust thanks to the

"Kindertransport," also survived being hit by a bus and spending many weeks near

death in a London hospital, this just two weeks before he was to go to Austria to

receive the Medal of Honor in June ’99. Having the ceremony essentially brought to

Peter; however, made it all the more significant, since his colleagues and friends

were able to participate and wish him well on this great occasion.

ALUMNI EMAILS

conduit! 10

tion product intended for ATMs, called“R3.” The latter ultimately resulted in animpressive product demo; however, as itturned out, R3 was basically ahead of itstime and was suspended.

While things did not work out as I hadplanned, I’ve since joined Sensar’s parentcompany, Sarnoff Corporation, working inthe newly formed Robot Vision Groupunder the Vision Technologies lab. (I didmanage to spend a week in Paris beforechanging jobs, a wise move!) The labworks on several government and com-mercial consulting contracts, and is alsovery interested in generating spinoffs (ofwhich Sensar was the first). So, I mayfind myself involved with a startup yet!Best wishes, Ted [email protected]

JACK HUMPHREY, ScB ’95Hello, conduit! readers—The last fiveyears have been great, both personallyand professionally. I am now living inAustin, Texas, and having an amazingexperience at Reactivity—more on thatlater.

After graduation, I joined Electronic BookTechnologies in Providence and helpedcreate a web content management systemcalled DynaBase. Jeff Vogel was my man-ager, and Ed Bielawa joined us at thebeginning of 1996—it was a great team!During my time at EBT (through July1997), we were acquired by Inso Corpora-tion, which has since downsized toinclude only the DynaBase business andis now called eBusiness Technologies. I’mhappy to report that the product lives on,is still evolving, and continues to be soldtoday.

When a small but significant blizzard hitNew England on April 1, 1997 (a day Ihappened to be scheduled to fly out to SanFrancisco), I admitted to myself that thisTexas boy was not cut out to be a NewEnglander. With nothing but the greatestadmiration for those of you who stand upto winter year after year, I headed back towhence I came, the sunny climes of Aus-tin. Here, I found a unique and ambitiouscompany called Trilogy and joined up.During my three years there, I helped todevelop infrastructure technologies forour e-business products, including a webapplication development framework that

only should you bring a lot of cash but youshould also plan well in advance becausechances are that several million NewYorkers are trying to do exactly the samething. For example, don’t just show up atthe door of the US Open without ticketsand expect to get in. And forget about get-ting seats for the Subway Series!

Here at IBM we are starting a projectcalled Blue Gene. Our goal is to build acomputer with a million processors and 1-petaflops performance (this is 100 timesmore powerful than the new ASCI White)to study the folding of human proteins.This is one of the most interesting ques-tions in biology. The function of a proteinis determined in large part by its three-dimensional structure, and many dis-eases like Alzheimer’s and mad cow dis-ease are caused by misfolded proteins.Understanding and simulating this pro-cess can help us provide new treatmentsand design new drugs. These simulationsare very compute-intensive, and a mach-ine as powerful as Blue Gene will stillrequire about a year of compute time tofold one protein, a phenomenon that canoccur in nature in just a few milliseconds.

Taking advantage of Blue Gene’s massiveparallelism is our challenge. Blue Geneshould still run in case of faults since weexpect a chip to fail every day. Our algo-rithms should provide predictable perfor-mance in this degraded machine. We willalso consume all the available powernorth of New York City but we should beable to slow down the machine in thesummer to cope with power shortages.

As you can see, I am trying to make gooduse of my Brown education. Please, don’tforget to mail me my conduit!. And, bythe way, I like the new web site much bet-ter. José [email protected]

TED CAMUS, PhD ’95Greetings! Well, after almost four excitingyears working on iris-matching algo-rithms and products, Sensar has beenacquired by IriScan, the company fromwhich we licensed the core iris-matchingsoftware. As you may remember, I hadworked on the WFOV (“wide-field-of-view”), which located the subject’s eye in3D space using stereo and face-templatematching algorithms, and was later thealgorithms lead for on our next-genera-

conduit! 11

typing new ideas for technology compa-nies. When these ideas take off, they canspin out into their own companies, withReactivity employees leaving as foundersand first employees. Zaplet (formerlyFiredrop) is our first spinoff and is doinggreat. I encourage everyone to check outtheir unique technology, which enablestruly dynamic email. Next time you wantto schedule a group meeting or poll yourfriends, go to http://www.zaplet.com.

I recently visited Brown on a recruitingtrip and got to visit with ProfessorsZdonik, Savage, and van Dam, as well asa few students. It was great to catch upwith everyone and hear about all of theinteresting things that are going on inBrown CS. I’d love to hear from anyoneout there in the extended community—feel free to drop me a line [email protected].

Shriram Krishna-murthi was born andraised in Bangalore, abeautiful cantonmenttown 3000 ft. above sealevel known as the Gar-den City of India. In1989 he came to the USon a scholarship toOhio Wesleyan Univer-sity in Delaware, Ohio.After Bangalore, Dela-ware seemed a staidplace to Shriram. Still,he enjoyed his timethere, jokingly callinghimself an academicquarterback because hesaw his hard workenhancing their GPAprofile!

For his junior year he took a semester offto visit Hungary and attend the BudapestSemesters in Mathematics program. Dif-ferent countries produce scholars whoexcel in different branches of mathemat-ics, and Hungarians are brilliant at com-binatorics—a prime area of interest to

computer scientists. His classmates thereincluded students from several Americanand Canadian universities. He knew inhigh school that he wanted to attendgraduate school, so he focused on a liberalarts curriculum at Ohio Wesleyan withthe agreeable anticipation of studyingcomputer science in grad school. He feelshe could not have received as broad aneducation in India.

Shriram did his graduate work at RiceUniversity because the CS faculty theretaught algorithms and programming lan-guages of just the right flavor to interesthim. His first year was spent on computa-tional biology with Alejandro Schäffer,working on the software package SOFT-LINK. During this period he co-authoreda paper on genetic linkage analysis,which has proven extremely populargiven the recent focus on the human ge-nome project. However, despite his suc-cess in this area, it was programming lan-guages in which he saw true beauty. SaysShriram, “If it’s not beautiful, you may aswell go and get a high-paying job!”

While Shriram was still trying to come upwith a suitable research topic for his the-sis, his advisor Matthias Felleisen had anepiphany that was to change both theirlives. Flying home from a conference on

supplemented ASP, Java Servlets, andJavaServer Pages (JSP).

Now I’ve been at Reactivity for two and ahalf months and I love it! (The only thingwe’re missing is other Brown grads.)Reactivity was founded in 1997 in SiliconValley, with the mission of bringingtogether technical innovators to buildgreat new companies. We now have officesin Silicon Valley, Austin, Boston, andSeattle. There are two parts to our busi-ness: startup services and startup cre-ation. We provide high-value technologyservices to our clients, usually fledglingtechnology companies who don’t yet havedevelopment teams. In a sense, webecome a client’s “virtual” product devel-opment team and build either a prototypeor a version-1.0 product. That’s startupservices, which is very rewarding andinteresting, but there’s also startup cre-ation—internally, we spend a lot of timebrainstorming, investigating, and proto-

TWO NEW CS FACULTY

ShriramKrishnamurthi

conduit! 12

related areas, which required studyingthe semantics and implementation ofadvanced programming languages, par-ticularly from the viewpoint of softwareconstruction. His work focused on how tocreate programming environments thatlet users build programming support fordomain-specific languages, how to buildlarge systems with combinations of these,and how to let the pieces written in eachlanguage communicate with one another.

Four years ago Shriram’s group at Ricestarted a high school outreach programcalled the TeachScheme! Project thattrains high school and college teachers inthe TeachScheme! curriculum. Whilethey’re still collecting statistical evidenceon the project’s outcomes, it appears to beespecially successful at retaining andmotivating female students. Next sum-mer, Shriram and his wife Kathi willteach joint high school workshops atBrown and Worcester Polytechnic. Shri-ram and his colleagues have produced atextbook How to Design Programs andhave developed the DrScheme program-ming environment. The three facets of theproject bring together the unique combi-nation of software design, a textbook thatlays it all out, and a programming envi-ronment that reinforces the principlesand grows with the students. Coinciden-tally, because both the textbook and soft-ware are available gratis on the Web, anincreasing number of home-schoolers,retirees and others have been using thismaterial.

Shriram is still getting used to the scaleof Providence; it’s small, but he has beenenjoying the architectural details of thebuildings, both old and new, and revelingin a city where he can really walk (hardlypossible in Houston). He has become aloyal WaterFire buff, drawn to it by itscommunal feel. He is looking forward tothe cold weather and to traveling upnorth, and is eager to do some kayaking.He is delighted that Providence is locatedat the convergence of two rivers, theWoonasquatucket and the Moshassuck,since he maintains that “any city worthits salt must have at least one river andone hill.” For the latter, College Hill cer-tainly fills the bill.

Shriram also enjoys the theatre and PeterGreenaway movies. He and Kathi thor-

education, he began analyzing his experi-ences. What, he wondered, was the singlemost frustrating aspect of educating stu-dents? They are all very smart, but decid-edly miseducated in computer science. Hedecided to change his entire researchfocus to teaching programming design tofreshmen; but he quickly realized that bythen it’s already too late, so studentsmust get their grounding in high school inorder to perform adequately as freshmen.A more modern and sophisticatedapproach to CS was needed. From thisemerged a new curriculum, Teach-Scheme!, and a plethora of pedagogicalproblems and a corresponding number oftopics for Shriram’s thesis.

Students need software tools that grow insophistication along with the students.Their implementors would like to reducethe construction burden by building theseextensibly. Components of his thesisincluded software construction and pro-gramming languages, two strongly inter-

This cartoon was a gift to Shriram fromDave Archer, one of the teachers in the

TeachScheme! program

conduit! 13

oughly enjoyed the RI Film Festival dur-ing their first week in town, seeing sevenfilms in four days. They’re still checkingout vegetarian restaurants in the areabut have found the Garden Grill to bebest so far. Since they like long road trips,they’re anticipating some major explora-tions of New England. He’s a seriousblues enthusiast and is currently lookingfor the best local clubs—all suggestionswelcome.

Michael Black was bornin North Carolina butgrew up in Baltimore.When he was 15, hisparents retired andmoved to Point Roberts,Washington, a smallpeninsula just belowthe 49th parallel sur-rounded by gloriousscenery, but devoid ofschools! He actually at-tended high school justacross the border inCanada.

He went to college atthe University of British Columbia inVancouver, starting out in psychology. Inhis second year he took a CS course,mainly because he knew he should, des-pite being terrified of computers: he feltthat computers were the future, andbesides, it was the only class that fit hisschedule. The course was programmingfor non-majors, mainly forestrymajors, and they learned Fortran pro-gramming on punched cards. Michaelfound this surprisingly appealing; heloved the tactile and aural sensationsof the cards going through the reader.His goal at the time was to become anarchitect, and there were plenty ofcourses on ‘computer architecture’around—unfortunately, of course,these were courses on the architectureof computers, not buildings! However,he found he could combine his interestin computers and humans by doingartificial intelligence.

After graduating he found there wereno jobs in Canada, so he went to Califor-nia hoping for a job in AI research. Withonly a bachelor’s degree, he was fortu-nate to land a job working on expert sys-

tems at GTE, then later doing computervision at ADS. Both companies paid forhis master’s degree at Stanford, which heworked on part-time. During this periodhe learned a lot about the corporateworld. He then decided on a PhD andwent to Yale, which was quite relaxing incomparison to holding down a full-timejob and attending graduate school atStanford. There, he worked with a groupof psychophysicists who were studyinghuman visual perception. He came toappreciate interdisciplinary research thatcombines computer science with neuro-science, cognitive science, and mathemat-ics. One of the things that attractedMichael to Brown was this same mix ofdisciplines and the collaborative spirithere, and he is excited to be part of theBrain Sciences Program at Brown.

When Michael graduated in ’92, he took apost-doctoral position at the University ofToronto and thoroughly enjoyed the expe-rience. There he met Allan Jepson, whoproved to be a great collaborator, sup-porter and friend. His next job took himto Xerox PARC. It was an appealing moveas PARC had always seemed a magicalplace where scientists, anthropologists,engineers, and artists mix with theenergy of Silicon Valley. Two years laterhe was managing the image-understand-ing area and eventually founded a newgroup in digital video analysis. Michaeldesigned this as an academic groupwithin an industrial lab with only a few

Michael Black

Behind the scenes at Michael’s photo session:Our attempts at an all-bubble background metwith limited success, despite valiant blowing by

Jennet Kirschenbaum and Dawn Nicholaus!

conduit! 14

Traveling and hiking are favorite pas-times, as are theatre and film. While liv-ing in Point Roberts, he worked at a statepark right on the water in the Strait ofGeorgia, where he enjoyed setting crabtraps and watching killer whales, whocame so close you could almost touchthem. He commuted by bike in Californiaand plans to continue biking in RI. Hefeels so passionate about his work that it

permanent researchers and many stu-dents, visiting faculty and sabbatical visi-tors, a new model and a more dynamicenvironment than was common at PARC.In the final analysis, however, Michaelwanted to be in an academic environment,where ten years along there would besome guarantee that he could pursue theresearch problems that interested him.

Nikolaos Triandopoulos (left in photo) was born and raised in Athens; both he and his par-ents had always hoped he’d continue his education in the US. Nikos developed his love ofcomputer science in high school, which led him to study at the University of Patras in theDepartment of of Computer Engineering and Informatics. After graduating in ’99, he spenta year in the math department at the University of Athens before applying to U.S. schools toget his PhD. The Kanellakis Fellowship was the deciding factor in his coming to Brown.Nikos enjoys traveling abroad, hiking in the countryside and skiing; he is looking forward toheading north to ski this winter.

Aris Anagnostopoulos was born in Houston, TX, where he lived for four years before his par-ents returned to Greece. Aris grew up in Thessaloniki, then moved to Athens when he wasnine. His interest in computers began as a hobby, when he would spend hours playing on hisfather’s computer. Like Nikos, he attended the University of Patras, and although he was ayear behind Nikos, they attended some classes together. Aris enjoys several kinds of music,both Greek and foreign, and has a huge CD collection. He also loves traveling abroad andmovies—preferably European ones.

Both Aris and Nikos had heard about Paris and his work from faculty members when theywere still at the University of Patras, so becoming Kanellakis Fellows was a particularhonor for them. Before leaving for the US and Brown, they visited General and Mrs. Kanel-lakis and were touched to see so many photographs of Paris and his family in their home.They plan a return visit over the Christmas holidays. Said Nikos, “We would prefer thatParis were here, of course, but we’re honored to have been awarded this fellowship.”

conduit! 15

was joined in his presentation by BobMonroe of FreeMarkets.com, a companythat conducts online auctions of every-thing needed by businesses including ser-vices. Bob gave a demo of an onlineauction in which suppliers across theglobe competed in real time to provide apiece of heavy equipment to a domesticcompany. It was fascinating to see compa-nies compete by progressively loweringtheir prices until the auction closed.

The next speaker was Steve LeBlanc ofCompaq, who addressed the topic of“Secure Enterprise Infrastructure Plan-ning for the New Millennium.” Hepointed out that (i) security breaks arecommon (64% of polled organizationsreported a break-in during the past 12months), (ii) security threats are rising,and (iii) security is a key obstacle to thesuccess of E-commerce enterprises. Hispresentation analyzed technologicaltrends for providing a secure networkinfrastructure for enterprises that areimmersed in the heterogeneous Internetenvironment. Various scenarios were dis-cussed that emphasized the pros and consof emerging security technologies, includ-ing: IPSEC, SSL, Windows 2000 securityfeatures, Common Data Security Archi-tecture and Public-Key Infrastructure.The talk concluded with a “how to” ondeveloping an enterprise security system,giving attendees a starting point fordeveloping their own strategic securityplan.

The third speaker was Jeff Kephart ofthe T. J. Watson IBM Research Cen-ter, whose talk was entitled “SoftwareAgents and the Information Economy.”Kephart envisions a future in which theworld economy and the Internet willmerge, and together evolve into an infor-

is almost a hobby as well as a job. He hasmade some excellent friendships andloves being part of a research commu-nity, enjoying the social network as muchas the science.

He has been married for 16 years to LeeMillward, who grew up in Montreal.They met in high school when Lee was asenior, about to leave for university. Shewasn’t sure she wanted to date a guystill in high school! Lee is a novelist so

her time is her own and she can accom-pany Michael on most of his conferencetrips here and abroad. They share thesame hobbies and are both excited aboutthis move to Providence. As a writer, Leefeels Providence offers a rich intellectualand artistic environment. They find thepace of life slower here than in Califor-nia, less frenzied and money-centric andmore balanced.

Last semester, the Department’sIndustrial Partners Program spon-sored its 25th symposium on theubiquitous topic “E-Commerce: ARevolution in the Marketplace.” E-commerce is pervasive in both thebusiness world of our industrialpartners, and the research world ofcomputer scientists and other in-formation technologists. The sym-posium encapsulated aspects ofbusiness-to-business (B2B) andbusiness-to-consumer (B2C) E-commerce, as well as agent and

other enabling technologies.

Steve Resnick of Microsoft kicked offthe event, speaking about “Business-to-Business Market Innovations.” Resnickpresented an overview of the changestaking place in the electronic market-place and the enabling technologiesbehind the scenes. Specifically, he des-cribed four transactional pricing mecha-nisms for marketplaces: forward auc-tions (many buyers, one seller), reverseauctions (many sellers, one buyer),dynamic exchanges (many buyers, manysellers), and electronic procurement (onebuyer, one seller). In addition, he empha-sized how the increased degree of B2Bconnectivity made possible by the Inter-net is affecting relationships among buy-ers, sellers, and their intermediaries:small-scale buyers and sellers can nowparticipate in large-scale global marketsthat were inaccessible to all but the larg-est of enterprises just a few years ago. Toprovide an example of a global market-place open to small businesses, Resnick

THE 25th IPP SYMPOSIUM

Host Amy Greenwald

conduit! 16

mation economy, bustling with billions ofeconomically motivated software agentsthat exchange information, goods, andservices with humans and other softwareagents. His talk surveyed research con-ducted by the Agents and Emergent Phe-nomena group at IBM on collective in-teractions among agents that dynami-cally price information goods or services.Among those topics surveyed, he des-cribed in some detail a model of shopboteconomics in which shopbots (compari-son shopping agents) strategically pricetheir information services, buyers tradeoff as to whether or not to use shopbots,and sellers use pricebots to dynamicallyprice commodities in response to pre-dicted buyer behavior. He presented sim-ulation results regarding the dynamicsthat arise in this model, demonstratingbehaviors ranging from price warsamong sellers to nonlinear pricing sched-ules for shopbots. These studies raisemany fundamental issues, both theoreti-cal and practical, particularly in therealms of multi-agent learning anddynamic optimization.

Just before lunch John Piescik ofAmerican Management Systemsspoke on the topic “Can Dinosaurs Learnto Fly? Big Brick-and-Mortar Compa-nies Don’t Necessarily Face Extinction.”

John’s thesis, as suggested by his title,was the competitive and institutionalchallenges of evolving companies in “In-ternet time,” offering hope for thosecould-be business dinosaurs. He pre-sented a set of dinosaur survival strate-gies that enable companies not only tosurvive but to prosper in a worldreshaped by E-commerce.

After lunch, Don Stanford of GTECH(and now newly appointed Adjunct Pro-fessor (Research) in the Department)spoke on the topic “Past and FutureTrends in High Speed Transaction Pro-cessing.” The talk focused on some of themore popular on-line transaction process-ing (OLTP) architectures, including theone employed by GTECH in its lotteryapplications. Don emphasized that futureaccess trends, such as the Internet andinteractive television, are causing trans-action processors to rethink current mod-els. For example, it should be possible totake advantage of the high penetration ofmobile phones in the consumer marketwith the recent developments regardingthe wireless application protocol (WAP),which add security and ease of implemen-tation. The future requirements for OLTPwere discussed in light of these newmethods of consumer access.

l to r: Steve LeBlanc, Compaq; John Savage, Brown; Boris Putanec,Ariba; John Piescik, American Management Systems; Steve Resnick,

Microsoft; Amy Greenwald, Brown; Don Stanford, GTECH; Bob Monroe,FreeMarkets.com; Jeff Kephart, IBM; Ed Gottsman, Andersen Consulting

conduit! 17

Roger Blumberg.In addition to his CS92 preparations,Roger Blumberg is teaching a course ti-tled “Science and Society in 20th-CenturyAmerica” in the History, Philosophy andSocial Sciences Department at the RhodeIsland School of Design (RISD).

Tom Dean.The department has been the recipient ofsignificant corporate largesse recently.Macromedia has given us software for ourNT machines to the tune of $150K—45copies each of Dreamweaver, Director, Au-thorware and Freehand, and six copies ofFontographer, as well as 90-day licensesfor copies of Web 101, Macromedia’s cur-riculum for using the above products, foruse solely in our new NT lab.

Amy Greenwald.In addition to participating in the TACcompetition and co-chairing last se-mester’s IPP symposium (see related arti-cles in this issue), Amy co-organized a

workshop on Multi-Agent Learning atICML in Palo Alto in July, and a confer-ence on Probability, Conditionals, andGames in New York City in August. Shealso presented a paper at Games 2000 inBilbao, the first international meeting ofthe Game Theory Society, whose mem-bership includes about 15% (andgrowing) computer scientists. The mostexciting of Amy’s summer escapades,however, was her participation in theCRA’s Distributed Mentor Program,through which she supervised twowomen undergraduates, Rebecca Hutch-inson and Gunes Ercal, who visitedBrown from Bucknell and USC, respec-tively.

Philip Klein.Philip is on leave from Brown andworking as chief scientist at a mobile-commerce startup in Berkeley. He hasthereby joined the ranks of those eager tohire Brown CS grads!

The next speaker, Ed Gottsman ofAndersen Consulting, spoke on thetopic: “Privacide—E-Commerce Opportu-nities in the Coming Panopticon.” Eddrew his inspiration from the Panopticon,a novel building design proposed in thelate 18th century by the philosopher Jer-emy Bentham to address the surveillanceproblems inherent in the management offactory workers and prisoners (Benthamdidn’t really discriminate between them).The Panopticon’s hub-and-spoke designand many one-way mirrors would let asmall number of supervisors secretlykeep watch over a much larger number ofsupervisees. Newly developing technolo-gies for the inexpensive acquisition,transmission, analysis, and delivery of

real-world data are creating the frame-work for a “collective Panopticon” inwhich everyone knows everything abouteveryone else. “Privacide” explores thetechnological and social forces working tobring about this modern panopticon, andthe opportunities it will create for elec-tronic commerce.

The final speaker of the day was BorisPutanec of Ariba (AB ’92, ScM ’93). Heentitled his talk “XML and B2B: TheBirth of a Standard.” In his talk Borisdescribed the process that Ariba has fol-lowed in creating an XML-based commoninteraction paradigm between companieswishing to engage in e-business transac-tions.

▼▼▼

▼▼▼

[email protected]

▼▼▼

conduit! 18

this field from physics and computer sci-ence. He chaired the international com-mittee for the award of the Gödel Prize,the premier award for outstanding andinfluential papers in theoretical computerscience (conferred this year on M. Vardiand A. Wolper). As the principal investi-gator of a four-member team including EliUpfal, he received a two-year $850Kgrant from NSF for computational biologyresearch, jointly awarded by the CISEand Biology Directorates. He has been re-cently appointed for a three-year term tothe Review Committee for the Mathe-matics and Computer Science Division ofArgonne National Laboratory.

Steven Reiss.Steven’s garden produced lots of egg-plants, peppers, and tomatoes in seasonand is still producing carrots, broccoli,pumpkins and Brussels sprouts. His cur-rent research involves completing variousaspects of his visualization system andexploring a new environmental frame-work to manage the evolution of softwaredesigns and code simultaneously. In addi-tion, he has a part in the ITR grant that

Shriram Krishnamurthi.Shriram was on the program committeefor the International Conference on Func-tional Programing, 2000, held in Montrealin September.

David Laidlaw.David and three students attended thevisualization conference in Salt Lake Cityin October. The students each presented apaper. He will be going to NIH in Decem-ber to talk about medical imaging work.He just received a prestigious NSF ITRgrant—$2.3 million over four years;details in an article above.

Franco Preparata.Franco presented a paper at Recomb2000,the foremost international conference incomputational biology, held this year inTokyo. In June he organized in Torino,Italy, for the Italian Accademia dei Linceian international forum on quantum com-puting that brought together leaders in

▼▼▼

▼▼▼

The “Sunlab” got a top-to-bottom overhaul this year. The populartiered classroom now sports Sun Ultra 10 workstations with 18” flat

screens mounted on articulated arms http://www.cs.brown.edu/system/hardware/workstations.html

▼▼▼

conduit! 19

was awarded to Stan Zdonik for studyingdata centers and Internet data access. Fi-nally, he is preparing to lead anotherjunior faculty search and to help in theprocess of upgrading or replacing ourcomputer equipment next summer.

John Savage.In May John was reelected to a third yearas President of the Faculty Club Board ofManagers. On October 5 the Club spon-sored a “Faculty Shelving Party” toreceive books donated by members of theBrown faculty for the newly renovatedClub Reading Room. Martha Joukowsky,President-Elect, who with her husbandArt made a major gift last spring toBrown to redecorate and furnish theReading Room, proposed that this partybe held to collect books for the new cabi-nets installed during the redecoration. Asmall crowd of very pleased faculty mem-bers attended and presented their booksto the Club.

Roberto Tamassia.Roberto has been appointed Director ofthe Center for Geometric Computing. Thesecond edition of his textbook Data Struc-tures and Algorithms in Java (coauthoredwith Michael Goodrich) was published inAugust by Wiley.

Eli Upfal.In August Eli was an invited speaker atthe Oberwolfach meeting on efficient al-gorithms in Germany. Together with hispostdoc, Milos Hauskrecht, he presenteda joint paper in the Fifth InternationalConference on Artificial IntelligencePlanning and Scheduling, in California inMay. The same month he visited theWeizmann Institute in Israel.

Andries van Dam.Andy was awarded the 2000 SIGCSEAward for Outstanding Contributions toComputing Sciences Education (“as ahypertext pioneer and a champion ofcomputing education for many years”)and was keynote speaker for SIGCSE2000. In addition, he was keynote speak-er for IEEE VR2000, and has initiated aproject that is brainstorming means ofestablishing a consortium of companies,government agencies, and foundations tofund a Grand Challenge-style set of inter-disciplinary research projects in educa-tional technology. He has been elected afellow of the American Academy of Artsand Sciences.

Pascal Van Hentenryck.Pascal had a frantic summer. He receivedan EEC grant in June and had to learnEuropean and Belgian budget proceduresin six days and seven nights (they don’thave Trina to help there). He then gavean invited tutorial at the ComputationalLogic Conference in London, finished apatent application, and flew to Atlantafor the International Symposium onMathematical Programming, where hewas cluster chair for constraint program-ming, before coming back to Providenceto welcome his family and prepare for hiscomputer architecture class. He also be-came associate editor of OperationResearch Letters in August.

Stan Zdonik.Stan received a $3.2-million, 5-year NSFITR grant for “Data centers: managingdata with profiles”, which will exploreusing profiles of users’ information inter-ests to manage web data intelligently;details in an article above.

▼▼▼

▼▼▼

▼▼▼

▼▼▼

▼▼▼

▼▼▼

▼▼▼

conduit! 20

Those of you who visit the CS web site areaware that the Sciences Library is beingfrequented by some peregrine falcons.Peregrines eat pretty much just other

birds, and the pigeon pop-ulation hereabouts hasthinned out considerably.We were finding randomwings and other non-delectable body parts onthe CIT decks. While inprinciple I can appreciatethe thrill of seeing one ofthese rare birds, I find thepractice much less inter-esting. Every time I haveseen them they look likesmudges sitting on the14th floor. On one occa-sion when editor-in-chiefSuzi Howe and a new fac-ulty member, ShriramKrishnamurthi, were en-thusing about the bird,the curmudgeon in me

came out with “When you’ve seen oneduck you’ve seen ’em all.” (Subsequently, Iwas reading a biography of Murray Gell-Mann, who is best known for postulatingand naming quarks. In the book Gell-Mann is described as somewhat of anintellectual showboat. Once a colleaguewho was hiking with Gell-Mann foundhimself forced to listen to Gell-Mann givethe Latin name of every bird they encoun-tered. The colleague re-sponded, “They all looklike ducks to me!”)

Elsewhere in this issueyou will see mention ofthe taping of the docu-mentary “2001: HAL’sLegacy” that I was in-volved in. As noted there,I assume that I wasasked to be the talkinghead who would deliverthe bad news about howfar current computer un-derstanding of languagefalls short of the “predic-tions” in the movie. Of

course, this is hardly news to any of us inthe field, and thousands of others couldand would have delivered the same opin-ions. I don’t know why I was chosen, un-less they were looking for someone with abowtie. At the end of the taping DavidStork (who is doing the interviews) askedme if there was any question I wished Ihad been asked. I said that a good onewould have been to ask me about wrongturns in the research path we have beentaking. So David asked me, and I saidthat I thought I had taken a 20-yearwrong turn. As grad students, I and mostof my colleagues all thought that it waspointless to worry about getting comput-ers to learn English, and that instead weshould try to program in the abilitydirectly. The argument is that since weare so much smarter than computers, ifwe can’t figure out how to do it, whatchance would a computer have? I nowbelieve this argument is wrong, despitehow logical it seems. It turns out that thecomputer’s speed, and in particular itsability to look at a lot of text, often cancompensate for its stupidity.

The department got some new carpetsthis summer. This was not announced,and I happened to be away the day theydid the fourth floor. The next day I wastreading on the thing when Jennet Kir-schenbaum (assistant to the departmentchair) asked me how I liked it. I had no

CHARNIAK UNPLUGGED

Eugene amid PBS’s recording gear

conduit! 21

idea what she was talking about. Eventhough the old carpet was green, and thenew one blue/purple, I had not noticed thedifference! There was even a “new carpet”smell that I only noticed after the fact. Iam really oblivious to my surroundings attimes. (Yes, I know, this paragraph wasn’tall that great—in fact, the editor told meso. But our editor-in-chief needs more ver-biage, and I also need to set up a line inthe next story.)

This one falls under the heading of “Prov-idence, beautiful Providence.” If you occa-sionally visit mid-town Manhattan, or area dedicated reader of the New YorkTimes (both descriptions fit me), you willknow that Manhattan has been invadedby a herd of plastic cows. The cows, aboutseven feet long by five feet high, are eachdecorated in an imaginative way, andgiven a fitting name. For example, a cowwith abstract heads and body partspainted on is called “PiCOWso.” I gathersuch plastic herds first popped up in Swit-zerland or some such place, and werethen copied in other cities, New Yorkbeing the most prominent. Providence,however, decided on something differ-ent—large replicas of Mr. Potato Head®.Why Mr. Potato Head? Because the

maker of the toy, Hasbro, has its head-quarters in Rhode Island. There is a Mr.Potato ATM in front of the Fleet Bankbuilding (though no money comes out)and a Ms. Potato Bishop in front of anEpiscopal church. At any rate, I was talk-ing with Suzi Howe when this topic cameup. Suzi deprecated them, and when Isaid I liked them Suzi pretended to retchon my carpet (the old one, we only got newones in the hallways). But Suzi’s reactionwas nothing compared to that of a neigh-bor of mine, a RISD faculty member, whoalso dislikes them. My potato defensebrought a look of such incredulity that Ithough I might be asked to turn in myRISD Museum membership card. Whatparticularly incensed my neighbor wasone stationed next to, and completelyupstaging, a Howard Ben-Tre sculpturenear the convention center. (Ben-Tre isprobably Providence’s most famous artist.One of his best sculptures can be found inthe main modern art gallery of the Metro-politan Museum in New York.) The Mr.Potato Head does, in fact, overshadow theconvention-center sculpture, but I thinkthis is the sculpture’s problem. The pieceshould have been placed indoors; outdoorsculptures have to be prepared to fight fortheir visual space, and this one is not up

Enough said!

continued on back page

conduit! 22

OU

R M

OST

REC

ENT

PhD

s

The new Microsoft lab (mslab) before its transformation

NAME ADVISOR THESIS NEW POSITION

CostasBusch

MauriceHerlihy

“A Study on DistributedStructures”

Assistant Prof. at Rensselaer PolytechnicInstitute working on distributed computing.“Recently, my research has been focused ontwo topics: counting networks and hot-potato routing. Counting networks arehighly distributed data structures used forefficient implementation of counters. In myresearch I improve on known counting net-work constructions and demonstrate howcounting networks can be adapted for sup-porting new kinds of operations. Hot-potatorouting is a bufferless form of routing whichis used in optical networks. In my research Icontribute new hot-potato algorithms for themesh network topology that improve signifi-cally on previously known algorithms.”

JoséCastaños

John Savage “Parallel AdaptiveUnstructured Computa-tion”

Doing research on the Blue Gene project atIBM T.J. Watson Research Center

VasilikiChatzi

FrancoPreparata

“Integer-CoordinateCrystalline Meshes”

Synopsys, Inc. in Mountain View, CA

Niyu Ge EugeneCharniak

“An Approach to Ana-phoric Pronouns”

IBM T.J. Watson Research Center

LeeMarkosian

John Hughes “Art-Based Modelingand Rendering for Com-puter Graphics”

Doing a 2-year postdoc at Princeton Univer-sity and collaborating with John Hughes ona project that they hope will result in a SIG-GRAPH submission this year.

SharonCaraballo

EugeneCharniak

“Automatic Construc-tion of a Hypernym-Labeled Noun Hierar-chy from Text”

Assistant Prof. at Georgetown University

conduit! 23

The 4th floor of our building was recently invaded by a 6-to 8-per-son TV crew and all their voluminous gear that was making a filmfor PBS showing next year, “2001: HAL’s Legacy.” Interviews withProfs. Andy van Dam and Eugene Charniak were taped for thisproject, which compares the achievements in computer scienceand technology portrayed in the 1968 movie “2001: A Space

Odyssey” with what has actually been done. WriterArthur C. Clarke and director Stanley Kubrick werewidely praised at the time for trying to ‘get the scienceright.’ How closely, the project asks, does the futurethey envisioned then match developments in informa-tion technology 33 years later?

What brought this project to Brown CS? Part of theanswer is that its writer, David Stork, had heard Andylecture some years ago and had also read his graph-ics text. It may also have been the intriguing contrastbetween the expectations 30 years ago and theachievements now in Eugene’s and Andy’s two fields.The ’60s were a heady time for AI and high hopeswere held for natural language processing—asembodied in HAL’s ability to communicate with all-but-human facility. Yet the intervening 32 years haveshown how extremely difficult such problems are. Incontrast, computer graphics, evident in the originalmovie only in primitive forms, has advanced beyondthe then state of the art, and is now well on its way towhat’s viewed as its next stage of development, full 3Dimmersive environments of the type simulated inBrown’s Cave.

Whether any of the Brown footage will survive the edit-ing process is something we won’t know, of course,

until the film is aired sometime in 2001. Until then, we look forwardto seeing our CS faculty in HAL’s company!

The mslab almost complete. Technical staffers JeffCoady and Tom Heft still working on the machines

“I’m sorry Dave, I’m afraid I can’t do that.”

Department of Computer ScienceBrown University

Box 1910, Providence, RI 02912, USA

conduit!

conduit! 24

NON-PROFIT

U.S. Postage

PAID

Providence, RI

Permit #202

Printed on recyled paper Address changes welcomed

conduit!A publication of

The Computer Science DepartmentBrown University

❦Inquiries to: conduit!

Department of Computer ScienceBox 1910, Brown University

Providence, RI 02912FAX: 401-863-7657

PHONE: 401-863-7610EMAIL: [email protected]

WWW: http://www.cs.brown.edu/publications/conduit/

Jeff CoadyTechnical Support

Katrina AveryEditor

Suzi HoweEditor-in-Chief

to it. However, I refrained from express-ing this opinion. I figured I was in deepenough trouble already.

Every year at graduation time there is aspecial faculty meeting for officially vot-ing to award degrees to that year’s recipi-ents. The meeting is usually pretty bor-ing, although it is not completely ceremo-nial. Every department sends a represen-tative armed with a list of people to whomit expects to award degrees. The facultymember must make sure that the listagrees with the registrar’s list, which isprinted in the agenda for the meeting.Occasionally mistakes are found. Typi-cally these mistakes are boring, althoughif you were an individual left out youmight not think so. This year, however,the registrar managed to come up with a

new and pretty racy mistake. In the list ofhonors and graduate degrees, the degreerecipient’s name is followed by the nameof the field in which the person is receiv-ing the degree (e.g., “Computer Science”).One such field is “Public Policy and Amer-ican Institutions”. This year, unfortu-nately, the registrar uniformly left the “l”out of “Public”. At the end of the meetingthe Pub(l)ic Policy representative stood tothank the parliamentarian, who hadbeen, interestingly enough, the only per-son to spot the mistake. I have saved acopy of the agenda as I figure it will be acollector’s item one day and also to fendoff accusations that this is an urban leg-end. The thought of what would have hap-pened if the degrees had been printed thisway still makes me smile.