25th ANNIVERSARY 79-0

4

Bro

wn

University Computer Scien

ce conduit! conduit!

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

Summer, 2004Volume 13, Number 125th Anniversary issue

Department of Computer Science Brown University

This article describes some of the lessons we havelearned from our collaborations with artists onvisualization problems. Over the past severalyears, we have worked closely with artists to de-velop, refine, and critique visualizations rangingfrom archaeological dig data from theGreat Temple of Petra site in Jordan to thefluid dynamics and wing bone shape defor-mations that begin to explain how bats, theonly flying mammals, stay aloft. Perhaps,the most important conclusion we havedrawn from this experience is that artistscan fill an important role in the visualiza-tion design pipeline. In our experience, art-ists routinely provide an unique source ofvisual insight and creativity for tacklingdifficult visual problems. They are also ex-pertly trained in critiquing and refining vi-sual works, an essential task in theiterative visualization process.

The second major conclusion we have drawn fromour collaborations with artists is that we needmore appropriate design tools to support them andtheir role. We discuss here the experiences that ledus to this conclusion along with some of the toolswe have developed to facilitate working with art-ists. The lack of appropriate design tools is partic-ularly evident in visualizations using new

technologies, such as virtual reality (VR) or vol-ume rendering. It is difficult for artists to get in-volved in design in these visual spaces since, withrare exceptions, one needs to know how to pro-gram in order to create within them. Unfortunate-ly, these are also the types of technologies thatoffer great potential for visualizing many of to-day’s complex datasets [4]. Additionally, they areprobably the technologies in which we can mostbenefit from artistic insight, since guidelines forgood visual depiction are far less developed in un-conventional visual spaces, such as virtual reality,than in more traditional 2D media.

We begin by describing one of our recent majorcollaborative efforts, a class on designing virtualreality scientific visualizations that was co-taughtwith professors and students from Brown’s com-puter science department and from the Rhode Is-land School of Design (RISD)’s illustrationdepartment. Many of the experiences and conclu-sions relayed here are the results of this class. Wethen discuss three important themes that we have

derived from our experiences, all motivated by adesire better to facilitate artistic collaborations. Insome cases, these themes can be thought of asguidelines for software tools that may aid collab-oration. In others we are not yet ready to offer aguideline, but we have at least identified issuesthat were major factors in our efforts and deserveconsideration before working with artists on visu-alization problems.

ARTISTIC COLLABORATIONIN DESIGNING

VR VISUALIZATIONS

artists routinely provide an unique source of visual insight

and creativity for tackling difficult visual problems...we need more

appropriate design tools tosupport them and their role

Daniel Keefe David Karelitz Eileen Vote David Laidlaw

conduit! 2

Teaching Art to Computer Scientists, Computer Science to Artists, and Fluid Flow to EveryoneOur interdisciplinary visualization class broughtartists and computer scientists together to solvevisualization problems driven by science. Stu-dents worked in teams on visualization and designassignments. We began the semester with 2D fluidflow visualization assignments, as in Figure 1,and gradually built up to the final projects, whichwere virtual-reality visualizations of pulsatileblood flow through a branching coronary artery.We found more obstacles to collaboration as wemoved towards VR and more complex data, asdiscussed below. Despite these obstacles, the stu-dents learned how to collaborate with one another,learned to value what each discipline (computerscience and art) could offer to the project, and pro-duced some very interesting visualizations.

Although artists rarely work with complex scien-tific data, they do train to convey information ef-fectively through imagery, given the constraintsimposed by their media, employers, or audience.In this abstract sense, normal artistic practice isnot such a far cry from typical visualization de-

sign tasks. The imagesin Figure 1 show one artstudent’s early visual-ization design assign-ment. We asked thestudents to create a vi-sualization and legendthat convey eight con-tinuous variables de-scribing a steady, 2Dfluid flow in a singlepicture. This is a verydifficult visual prob-lem; in fact, it is still be-

ing actively researchedin the visualization com-munity. We found that

artists were adept at investigating visual problemslike this one when we could clearly convey thescientific goals and constraints of the problem.

Collaboration was sometimes difficult to manage.In early assignments, such as in Figure 1, the righttools for the job were colored pencil, oil paint,gouache, watercolors, and Photoshop. In later as-signments, the essential tool for the job movedcloser and closer to programming. At this point,the art students often had visual insights to offerbut had difficulty conveying them. It was easy forthe non-programmers to feel left out of the loop.As Fritz Drury (the RISD illustration professorwho co-taught the class) remarked, the program-mers are the ones with the ultimate power: theyhave the final say about what ends up on thescreen.

One device that helped us keep artists, computerscientists, and fluid flow researchers on the samepage is the critique, a common teaching tool in artclasses. All the class work was displayed on awall, as seen in Figure 2, and as a class, we dis-cussed important design lessons in relation to eachwork. We critiqued the work both from a visualand a scientific standpoint. Visually, we exploredcolor, scale, form, metaphor, and narrative. Scien-tifically, we learned about the data we were tryingto represent and critiqued the work on the basis ofhow truthfully and completely the science wasrepresented, given the tasks our scientists wishedto perform. We have now adopted ‘crits’ into thevisualization development process for many ofour projects.

How Can Artists Approach Design Problems in VR?As we move from 2D visualizations into morecomplex 3D situations such as virtual reality, col-laboration with artists becomes much more diffi-cult to facilitate. The first theme we have derivedfrom our class experiences (along with other col-laborative efforts) is that visualization designshould occur within the visualization target medi-

Figure 1. An art student’s visualization design of 2D steady fluid flow past a cylinder. Courtesy of Deborah Grossberg

Figure 2. Students prepare for a critique of arterial blood flow visualization designs

conduit! 3

um. This sounds simple, but it has fairly signifi-cant ramifications for the visualization media weoften use. For example, it is very difficult for any-one, and nearly impossible for an artist who is nota programmer, to create visualizations or simplyexperiment with design ideas in virtual reality.

A starting approximation for designing within VRis to design with more traditional, often 2D, mediaand hope that some of these design ideas willtranslate to VR. We were forced to take this ap-proach during many of the class assignments. Thedifficulty is the drastic difference between whatwe can convey on paper and what we can conveyin VR. We use a four-wall Cave VR display envi-ronment for much of our research. So muchchanges when we enter the Cave: scale, interac-tion, stereo vision, vividness of color, and con-trast. When designing traditionally with an eye

towards VR, weare left with theproblem that agood 2D designdoes not neces-sarily translateinto a good 3D,much less VR,design. Further, itis very difficult toevaluate or pro-

pose refinements to a design without actually see-ing it implemented in the Cave. We thus lose thepower of the critique, which we have found souseful. We need to be able to design and critiquewithin VR.

With this motivation in mind, we began exploringways to work with artists to design visualizationsdirectly within the Cave. Figure 3 shows somesnapshots of one of our VR design results. In thisproject we are collaborating with Sharon Swartzof Brown’s evolutionary biology department, whostudies bat flight from experimental data collectedin wind tunnels. Two important clues to under-standing bat locomotion are the air-flow informa-tion surrounding the wing and the pattern ofdeformations of the wing bones during flight. Art-ists worked directly in VR to create the visualiza-tion design shown in Figure 3. Since the bat data

assumes symmetry between the two wings, theartists chose to represent different aspects of thedata on each side of the bat. On the left side ofthese images, flow close to the wing is describedby color and texture along the wing surface. Vor-tex cores and vortical structures in the flow behindthe bat are also represented. Changes in boneshape at two distinct times during a wing beat cy-cle are shown on the right side of the images alongwith a 3D trace of an important bone joint throughthe wing beat cycle.

The basis for our VR design tools is the Cave-Painting program [3], a tool intended for artists touse inside the Cave environment to create free-form 3D objects. It has been described as a formof zero-gravity sculpture. Artists interact with thesystem by moving a tracked paintbrush propthrough the air to create 3D ‘paint’ strokes. (Fig-ure 6 shows an artist using the system.) The‘paintings’ are actually 3D models, since eachbrush stroke exists in 3-space. The intuitive inter-face of the system makes it easy for artists to pickup and quickly begin modeling in the Cave.

There are several benefits to working directly inthe Cave with a tool like CavePainting. The mostimportant is that the design can be easily critiquedand refined with proper attention to the nuances ofthe target medium. In practice, we have gainedvaluable insight from these critiques. We havemade several alterations to our initial bat visual-ization designs based on feedback from Dr.Swartz and her collaborators after meeting for cri-tiques in the Cave. During these critiques we haveeven been able to sketch modifications to designsand discuss them immediately.

Using CavePainting to design visualizations alsohas the advantage that we can investigate, refine,and converge on a successful visual design at anearly stage in the process. With the usual approachof implementing before visual refinement, itmight take weeks or months of implementationbefore we discover our design is flawed from a vi-sual standpoint, and once we notice a problem andbrainstorm another design, it could take anotherfew weeks before we are ready to visually critiquethat one. Thus, particularly in VR, where imple-

Figure 3. A CavePainting visualization design of bat flight data, snapshots from a 3D VR program. The bat appears to fly into the page in these snapshots, but viewers walk

around the entire model when seen in VR

a good 2D design doesnot necessarily translate

into a good 3D, muchless VR, design

conduit! 4

mentation can be difficult and time consuming,putting visual design decisions in series with im-plementation can extend the time between itera-tions on a visual design. Designing directly in VR,on the other hand, lets us converge upon a visuallysuccessful design early in the implementation pro-cess. We can quickly work through many more it-erations of the design because we do not have towait for them to be implemented before critiquingthem in the Cave.

Where’s the Data?The second theme that has emerged from our col-laborative work is that we should incorporatevarying levels of data involvement in the designprocess. In the bat visualization design shown inFigure 3, there is no programmed link between thevisuals and the bat flight data. Designs such as thislie at one end of the spectrum shown in Figure 4.Despite the lack of a low-level link with the data,this type of design is extremely useful. The de-signer has imagined some representative data andsketched it out. The visualization is not farfetched; he has seen previous attempts at bat visu-alizations and talked with the scientist about hergoals. Essentially, he knows enough about thestructure of the data to sketch out a typical situa-tion so that we can meet with the scientist and cri-tique the visualization idea in the Cave.

The danger in going too far in the design processwithout a program-level connection of the visualsto the data is that we may converge upon a designthat works well for our perception of the data butnot so well with the actual data. In an effort to ex-plore this issue, we built some design tools to ex-plore the other end of the spectrum of datainvolvement. As we see in Figure 4, tools likeCavePainting lie at the far left end of the spectrum,with no program-level connection to the scientificdata. The visualization prototyping system de-scribed below is much closer to the right side ofthe spectrum, where data plays a key role in gen-erating the visualization design.

Our visualization prototyping system [2] lets anartist draw icon-based 3D visualizations that arecompletely driven by the underlying scientific da-ta. Figure 5 is a snapshot of one such visualizationdesign. The squidlike icons represent data valueswithin a fluid-flow dataset of pulsatile blood flowthrough a branching coronary artery. In this de-sign, the squid’s tentacles morph in response todata values. At high speeds, they straighten outand the squid appears quite streamlined. At lowerspeeds, they flail out to the sides, as the squid as-sumes a sluggish posture. This tool has been use-ful in evaluating several different designs forarterial blood flow visualization. Since we areworking with time-varying, pulsatile fluid flow,the ability to see the design animated, with iconsflowing down the artery and changing shape in re-sponse to the data, is critical in evaluating the de-sign’s success. This would be a difficult display torealize without a program-level link to the flowdata. Despite the success of this approach inachieving these animated visualization designs,we have had difficultly moving beyond these rel-atively simple cases to the more complex ones re-quired in many of our driving scientific problems.

This experience illustrates the tradeoff that existsin many design systems based on the role theyprovide for data. Given plenty of preprogrammedconnections to data, design tools can produce vi-sualization designs that are so representative ofthe data that they can be trusted and critiqued ascompletely accurate visualizations. However, pre-programmed connections to data can be constrain-ing to the artist. For example, in our currentimplementation of the prototyping tool, the iconsmust be drawn in a special way in order to estab-lish a solid correspondence for our morphing algo-rithm. This means that the artist must have this inthe front of her mind while working on the design.Creating very complex designs, for example iconsthat respond to six different variables, can becomealmost impossible to manage cognitively. Again,these difficult design tasks are the ones our drivingscientific problems require and the ones in whichwe can most benefit from artistic insight. We needto continue to develop intuitive design tools thatprovide this type of solid connection to the data,but also allow artists to work naturally.

Figure 5. A result from our visualizationprototyping system. The 3D icons were

sketched by an artist, then connected to an arterial blood flow dataset so they morph in

direct response to the data

Figure 4. Design tools can have a stronger or weaker built-in connection with data. Tools at

both ends of the spectrum are useful

conduit! 5

Needed: Tools for Real Design Work in VR

The final theme that has emerged again and againin our collaborations is the need to support contin-ued, evolving work with VR tools. This has beenevident in two areas. First, getting started in VR ishard. Often our artists have done several prepara-tory sketches or studies before entering the Caveto work. We need to make it easier for them tobuild on those sketches when they get to the Cave,rather than shutting out the real world and concen-trating only on VR. Second, we need to facilitate

sketches or paintings, directly into the VR designprogram. In Figure 6, we see a 3D CavePainteddesign inspired by a Miró painting [1]. One of ourdesigners saw the painting and it prompted an ideafor visualizing the bat dataset. We cut out subre-gions of the Miró and imported them into Cave-Painting as brush stroke textures. Then ourdesigner was able to work directly with elementsof the inspirational imagery to create the 3D de-sign she imagined. This gave her a jump-start onher 3D design and helped her quickly create a co-herent design.

The ability to return to a design and refine it againand again is just as important as starting withsomething in VR. The design task is necessarilyan iterative one, with critiques by other designers,implementors, and scientists all playing an impor-tant role in refining each iteration. Normally art-ists refine work in two ways. First, they addadditional layers of clarification. In painting, forexample, additional layers of paint conceal whatlies below. A rough outline of a face can be laiddown as a place holder for a much more complexrendering to come later, applied with additionalpaint layers. Second, they create many studies ofan idea, sometimes ending up with a studio full ofrenderings and re-renderings. At the end of thisperiod the idea is clear enough in the artist’s mindthat she feels ready to produce a final work.

These approaches are not at all mutually exclu-sive; however, we have difficulty supporting ei-ther with our current design tools. In the first case,we can add some additional layers of clarificationwith the CavePainting system, but this can havethe effect of distorting the original form. We are alittle closer to supporting the second style of re-finement, which amounts to letting an artist quick-ly reel off many sketches before creating a finalwork. However, it is unclear how to refer back toseveral studies while working on a new piece,since each design is usually intended to be viewedin the full space of the Cave. These issues areamong the most important to address before work-ing closely with artists on design problems, sincethey can be very frustrating and limit the amountof real design work that can be accomplished.

Conclusions

We have had many exciting and fruitful collabora-tions with artists, and we are convinced that theyhave a place in the visualization design pipeline.One of the driving motivations in our recent workhas been to consider what an artist would do foreight hours a day if hired by a visualization lab.Given current visualization practice, this is a dif-ficult question to answer. However, the key seemsto lie in enabling an artist to get involved in designat a level that goes deeper than simply turningknobs of existing visualization techniques. We an-ticipate that artists will be hired to fill positions inexploratory visualization. That is, rather than

Figure 6. The 3D visualization design, also for the bat flight problem, was inspired by the Miró painting “The Gold of

Azure” [1]. We scanned and imported the elements of the painting into our Cave design program as textures. The designer used this inspiration to hit the ground running

when she began her design in the Cave

returning again and again to a design to reworkand refine it. The real-world problems with whichwe anticipate artists will work are sufficientlycomplex that they will require many design itera-tions to complete. Tools to facilitate artistic col-laboration in visualization need to be accessible toartists in these ways if we want to support artisticinvolvement in difficult visual problems.

Let us look at CavePainting again as an exampleof an artistic design tool to see how it can be dif-ficult to get started on a VR design. When the pro-gram begins, we walk into the Cave, a dark blankroom of projection screen walls. We carry atracked paintbrush prop and a pair of glasses.Once we put on our stereo glasses, it is too dark tosee any paper or other real objects we might havebrought in with us. By default, we start with acompletely blank canvas and no external inspira-tion, something designers almost never want todo.

One approach that has been helpful to us is to im-port our design inspiration, often 2D work such as

conduit! 6

merely making a picture pretty or visually clearfor publication, we see artists as having key rolesin working closely with scientists to design novelvisual techniques for exploring data and testinghypotheses.

From our experience, three of the most importantthemes to consider in trying to obtain this goal are:

◆ Enabling artists to design directly with typical visualization target media.

◆ Considering and supporting the varying rolesthat data can play in artistic design tools.

◆ Facilitating getting starting and continuingto refine designs for difficult visualization problems.

We hope we have illustrated some of the potentialof this type of collaborative visualization work,along with presenting some of the lessons we havelearned along the way in our collaborations withartists. We also hope to have further motivated theneed for additional research in design tools thatcan be easily targeted towards visualization prob-lems.

AcknowledgmentsWe would like to thank the Brown and RISD stu-dents of our visualization design class, as well asRISD illustration professor Fritz Drury. We would

also like to thank our collaborators Sharon Swartzand Peter Richardson. This work uses the VRPNlibrary, which is supported by the NIH NationalResearch Resource in Molecular Graphics andMicroscopy at the University of North Carolina atChapel Hill. This work was partially supported byNSF (CCR-0086065).

References [1] Miró’s “The Gold of the Azure” may be seenat: http://www.bcn.fjmiro.es/angles/_coleperm/_salaanys6070/lordelatzur.html

[2] David Karelitz, Daniel Keefe, and David H.Laidlaw. Using CavePainting to create scientificvisualizations. IEEE Visualization 2003 PosterCompendium, October 2003.

[3] Daniel Keefe, Daniel Acevedo, Tomer Mos-covich, David H. Laidlaw, and Joseph LaViola.CavePainting: A fully immersive 3D artistic me-dium and interactive experience. In Proceedingsof ACM Symposium on Interactive 3D Graphics2001, pages 85-93, March 2001.

[4] Andries van Dam, Andrew Forsberg, DavidH. Laidlaw, Joseph LaViola, and RosemaryMichelle Simpson. Immersive virtual reality forscientific visualization: A progress report. IEEEComputer Graphics and Applications, 20(6):26-52, November/December 2000.

NEW FACULTY MEMBER, Assistant Professor John Jannotti joinedthe department this spring. He was born in Newburgh, NY, and isthe youngest of four children. In junior high he enjoyed chemistryand didn't use a computer until 11th grade—when he found thecomputer much more fun. He took computer classes but re-mained 'well-rounded' academically, and didn't become a realcomputer geek until college. John played varsity soccer and trav-eled widely with the debating team; said he, “I just liked to argue!"In high school John won a T.J. Watson Scholarship, in graduateschool he won an IBM Fellowship, and at Brown he says, "Now IBM'sbuilt my building!"

At MIT he minored in architecture, his interest sparked perhaps byhis father's having designed and built the family home. Duringgrad school he was involved in a .com called SightPath, foundedby his advisor Frans Kaashoek and David Gifford, that was eventu-ally bought by Cisco. His job was to write their prototype contentdistribution system. He worked for SightPath on and off until hegraduated. During a period as a post-doc at MIT, he worked forCisco on the same system before coming to Brown.

John lives on the East Side and likes it that Providence is geograph-ically small and walkable. His hobbies include growing and prop-agating house plants--he's hoping for a south-facing office on the

third floor. He plays soccer with the department's 'ByteSoccer' team, and softball with the'Dingers'. Recently he had his first flying lesson at T.F. Green Airport. John’s area of interestis computer systems, broadly construed, especially networking and operating systems. Heis particularly interested in loosely coupled distributed systems enabling qualitatively newfunctionality. Last semester he taught CS296-6, Large-Scale Networked Systems, and in thefall will teach his new course, CS161, Building High-Performance Servers.

conduit! 7

In March the depart-ment’s Women in Computer Sci-ence (WiCS) group held its annualGirl Scout Technology Day. GirlScout Troop 941 from Franklin,Mass. participated in order to earntheir technology badges. The scouts, ranging in age from 10to 13, first learned about the com-ponents of computer hardware, in-cluding input and output devices,types of memory, and the CPU.With the help of WiCS members,the girls worked in groups to dis-assemble old computers to see allthe components and how they fittogether to form a computer. Thescouts particularly enjoyed the“ripping apart” aspect of the exercise.Each scout took one computer part homewith them for show and tell. After a pizza lunch, WiCS members ledthe girls in a game of Hardware Bingo to

help them remember what they hadlearned. WiCS members followed thiswith an HTML tutorial, where the Girl

Scouts were able to createwebpages individualized withtheir favorite colors, theirchoice of images, and links tofavorite sites. These webpages,filled with information abouttheir pets, hobbies, and fami-lies, and pictures from the eventare linked off the WiCSwebpage, which can be found athttp://www.cs.brown.edu/peo-ple/orgs/wics/. Participating WiCS membersincluded Andi Fein, TeresaMcRann, Sara Hillenmeyer,Krista Greer, Caitlyn Schmidt,Casey Jones, Katrina Ligett,Stacy Wong, and Danielle Karr.Additionally, two students fromthe Artemis Project 2003,Dulissa Rosario and Ashley

Fernandes, joined WiCS members tohelp teach the material they learned inArtemis last summer. The Girl ScoutTechnology Day was funded by a recentgift to WiCS from Google. WiCS and thegirl scouts concluded the day with a dis-cussion of all the career opportunities intechnology-related fields, including jobsin teaching, programming, technical sup-port, and research.

Upcoming WiCS events include a talkby Professor Pascal Van Hentenryck, acourse registration bagel lunch, theannual senior brunch, and a tentativelyscheduled trip to New York City in thefall to tour Goldman Sachs’ TechnologyDivision.

WiCS holds GIRL SCOUT TECHNOLOGY DAY

Andi, Dulissa and Sarah pose withGirl Scouts and computer parts

Top l to r: Ashley, Casey, Teresa and Krista with their hardware group

Sarah shows Girl Scout Emily Lutherhow to add a link to her webpage

Ex-avd student Stephen Cantrill received the Outstanding Teaching Award at Denver Health’s annual medical staff dinner in October 2003. He is associate director of emergency services at the University of Colorado Health Sciences Center and also serves on the Colorado governor’s emergency epidemic response committee. His son Bryan and Mike Shapiro (both ’96) wrote an article in the fall 2002 issue of conduit! about their lives at Sun Microsystems.

------8<------

Coincidentally, Shriram Krishnamurthi noticed a glowing article about Bryan Cantrill (center in photo next page), Mike Shapiro (l) and Adam Leventhal (’01) in a British publication, The Register, under the heading, “Sun delivers

changelog

conduit! 8

LIDDY SHRIVER, ’90

In the spring ’03 issue, we learned of Liddy’s illnessand her great enthusiasm for biking and read her in-spirational online journal. It is with great sadnessthat we report her passing this January. From the fi-nal entry on her website:

January 15, 2004—Liddy took her final steps on herjourney with Ewing’s sarcoma today at 11:15 AM.No more pain. No more labored breathing. No moretrembling hands and wobbliness. No more auras.She is at rest and at peace. A version of the obituaryprinted in the local newspaper follows:

Shriver, Elizabeth:

Elizabeth Anne Marie Shriver, known to her familyand friends as ‘Liddy’, died on Jan. 15, 2004 aftera 22-month battle with Ewing’s sarcoma.

She was born November 12, 1966 in East Los Ange-les, CA, to Bruce and Beverly Shriver, who live inOssining, NY. Liddy is survived by her husband TomSwartz, her parents, and her three brothers Bruce,Jr., Mark, Matthew and their families. Liddy residedwith Tom in Jersey City, NJ.

She was a graduate of SUNY/Stony Brook (BS),Brown University (MS) and NYU (PhD). Sheworked at Lucent’s Bell Labs in New Jersey aftergraduating from NYU. Liddy was an avid cyclistand during her bout with cancer she inspired manypatients and caregivers with her online journal andher bike tours.

A graveside ceremony was held at St. AugustineCemetery in Ossining, NY, this spring and a memo-rial service took place at Liberty State Park in NewJersey, where Liddy and Tom frequently biked.

In lieu of flowers, the family requested that dona-tions be made to the Liddy Shriver Sarcoma Initia-tive, a non-profit charitable entity for those dealingwith sarcoma. Your check should be made out to‘FJC’ and the memo line should read ‘Liddy Shriv-er Sarcoma Initiative’. Please send the check toFJC, 520 Eight Avenue, 20th Floor, New York, NY10018.Peace,

Bruce, Bev and Tom

Unix shocker with DTrace.” DTrace is a key addition to Sun’s flagship

operating system in the upcoming release of Solaris 10. Bryan is quoted as saying, “With the exception of system calls, the [debugging] tools, such as they exist at all, are ad hoc, and at best designed for developer use. For example, there is no tool anywhere that allows for arbitrary dynamic instrumentation of a production operating system kernel.” To read this fascinating piece, go to: http://www.theregister.co.uk/

2004 07/08 dtrace_user_take/

------8<------

John Hughes made another good alumnus find in The New Yorker’s piece on “dorkbot”, a group whose

changelog

conduit! 9

Modern computer tools are becoming in-creasingly complex, from gigantic data-bases to computer-aided surgery. Yet aprimitive virus can still disrupt millions ofpersonal computers in one day. Two natu-ral questions to ask are what brought aboutthis sad state of affairs, and what can bedone to fix it?

Where do security problems come from?When testing software, one cannot possi-bly test for all cases and might miss a bughere and there; the bigger the system, the

more bugs creep in. Low-level architectureand operating-system design also affect se-curity. For example, as systems are de-signed today, an attacker can take controlof a program that has a buffer overrun bug:a buffer overrun can cause a program tostart executing malicious code. Even in theabsence of bugs and viruses, security prob-lems can arise as a result of a bad access-control policy or other unanticipated userdecisions, such as file sharing.

The CS Department recently hosted an IPPSymposium on the Trusted ComputingGroup’s (TCG) effort to make personal

computers more secure. TCG (www.trust-edcomputinggroup.org) is an industrialconsortium whose membership includesmost major computer companies, includ-ing several of our IPP partners: Microsoft,Sun, IBM, and Intel. The TCG replacedthe Trusted Computing Platform Alliance(TCPA) a year ago, incorporating all oftheir previous work.

The TCG is trying to create a personalcomputer that does not have the gapingsecurity holes present in today’s PCs. Italso aims at creating technology to allevi-ate security problems with platforms otherthan just the PC. This involves a separategadget—a trusted platform module(TPM) chip—that can insure that eachprogram’s data is protected, that programsdo not interfere with one another and thattheir code is not modified. A PC or otherdevice that has a TPM chip would be ableto prove to other machines on the networkthat it is enhanced with this technology.

The computer companies involved withthe TCG are working on all aspects of thissolution. They are creating standards formaking this chip and the supporting soft-

ware. Intel is making a TCG-compliant chip and shipping itwith its chipsets, and Microsoftis developing an operating sys-tem called NGSCB (“Next-Gen-eration Secure ComputingBase,” pronounced “ensgkib,”formerly known as Palladium)that uses TCG technology. AllTCG partners are involved inwriting the specifications.

On March 25th, some of the keypeople behind TCG specification designcame to Brown and presented their ideas.The title of the symposium was “TrustedComputing Group: Goals, Achievements,and Controversies.”

Joe Pato (HP), Brown ’81, kicked off theday by covering the “goals” part of theSymposium. He explained the currentstate of affairs that motivates the TCG ef-forts, and the vision for developing theTCG architecture. He explained that thecore idea of the TCG architecture is tobuild a large secure system by using onestandard hardware component that can be

changelogIPP SYMPOSIUM:TRUSTED COMPUTING

GROUP

Anna Lysyanskaya

The TCG is trying to create a personal computer

that does not have the gaping security holes

present in today’s PCs

motto is “People doing strange things with electricity.” At the meeting last May in SoHo, Scott (Spot) Draves (Math ScM ’90) was the first speaker. His presentation concerned his Web site, Electric Sheep, on which abstract images constantly change form in response to information coming from other people’s computers via the Internet. Said Draves, ”The project started in ’99. It was based on an algorithm developed in 1992. The title is based on the novel Do Androids Dream of Electric Sheep?, by Philip K. Dick.” Said he at the end of his talk, “Please buy my DVD. I quit my day job to do this.”

------8<------

Steve Marschner ’93, former CS224 TA and, according to Spike, “general all-around nice guy”, won a Technical Achievement Academy

conduit! 10

that a user can interact with an applicationwithout interference from other applica-tions. Finally, attestation allows a pro-gram to prove to another program,running on a remote machine, somethingabout its state.

Ernie Brickell (Intel) described howTPM version 1.2 was designed to supportthe cryptographic functionality needed fora trusted platform. He went over the pur-pose, functionality, and implementationof the cryptographic protocols. In particu-lar, he described the protocols for sealingsecrets, creating and using an endorse-ment key, creating and using an attestationidentity key, and resetting and extending aplatform configuration register and local-ity. The ability to seal secrets has manyuses; for example, it is crucial for makingdata available only to authorized applica-tions. Endorsement and attestation identi-ty keys come in, for example, when anapplication is required to prove that it is ina certain state. Each TPM has a public en-dorsement key (which can be thought ofas an identifier for this TPM) and a corre-sponding secret key. If a TPM also has anendorsement certificate from some trustedauthority (for example, the entity thatmanufactured this particular TPM), then itcan prove to third parties some propertiesabout the state of its platform. More pre-cisely, it provides a verifiable statement ofthe form “An entity in possession of a se-cret key corresponding to a certified pub-lic endorsement key attests that itsplatform is in a state with the followingproperties,” etc. From the privacy point of

trusted to perform a few simple tasks cor-rectly and securely. The goal of the stan-dardization effort is to establish what thesetasks are and to come up with a set ofbuilding blocks that make it possible se-curely to realize any type of computingplatform. Having one established industri-al standardization effort around is a goodidea when public review is concerned—the more experts are involved in critiquinga standard, the better. Also, anyone shouldbe able to produce products that complywith the standard.

Next, I gave an overview talk on the widerchallenges of creating trustworthy sys-tems, such as what constitutes identity inthe digital world (in my view, it is theknowledge of a secret key or password,since online you have nothing but yourdata to represent you), identity theft versusidentity fraud (the former is when some-one steals your identity, the latter is whenyou let your friends pretend to be you sothey can enjoy the same privileges you do)and how to prevent them, and other chal-lenges.

Brian LaMacchia (Microsoft) presentedthe Microsoft NGSCB operating systemwhose security features rely on the TCGstandard. The four key components of theNGSCB are strong process isolation,sealed storage, secure path to and from theuser, and attestation. Strong process isola-tion guarantees that one program cannotinterfere with another one’s execution.Sealed storage protects the data of oneprogram from another. Secure path insures

Award. The Oscar went to Steve and his collaborators for their pioneering research in simulating subsurface scattering of light in translucent materials. These ground-breaking techniques were used to create realistic-looking skin on digitally created characters.

------8<------

You’re sure to enjoy the highly entertaining Folklore Web site created by Andy Hertzfeld ’75, a student of Andy’s, one of the four legendary designers of the Mac, and an originator of Apple. In particular, check out the tale of the Mac team’s visit to Brown, entitled “What’s a Megaflop?”http:/www.folklor.orgStoryView.py?project=Macintosh&story=Whats_A_Megaflop?.txt

------8<------

CLASS OF ’22?...Gamze and Ugur Cetintemel’s baby daughter, Ece (Little Princess in archaic

changelog

Symposium speakers, l to r: Ari Schwartz, CDT; John Jannotti, Brown; Seth Schoen, EFF; Ernie Brickell, Intel; Anna Lysyan-

skaya, Brown; Jan Camenisch, IBM Zurich; Brian LaMacchia, Microsoft; Joe Pato, HP Labs; Simson Garfinkel, MIT

conduit! 11

view, it is important to insurethat this statement does notactually reveal the endorse-ment key or any other infor-mation that can uniquelyidentify the underlyingTPM. Otherwise, it would

be equivalent to providing your name andaddress to everyone with whom you wishto talk. So instead, this statement shouldreveal nothing except the properties of theplatform that are needed, and also someattestation identity key, which is just apseudonym that this TPM can use in fur-ther transactions with the same third party.

Jan Camenisch (IBM) presented the un-derlying cryptographic technique for thisprotocol (called “direct anonymous attes-tation”), as well as how to prevent “rogueTPMs” from making such statements. Thetake-home message is that you can haveauthentication and anonymity at the sametime.

A very interesting part of the day was thepanel discussion that followed these tech-nical talks. Besides earlier speakers, Sim-son Garfinkel (MIT), John Jannotti(Brown), Seth Schoen (Electronic Fron-tiers Foundation) and Ari Schwartz(Center for Democracy and Technolo-gy) participated in the panel. The audi-ence also joined in the discussion.

A number of controversies around theTCG standard were discussed. Both Sim-son and John raised pragmatic concernsover whether this effort was likely tomake an impact at all. They doubted thatpeople would write software that usedTCG or the security features of NGSCB;they felt that people would find thesethings too hard to use. Someone willing tomake a time investment to learn how towrite better, more secure code wouldprobably opt for studying better softwareengineering practices rather than the tech-nicalities of the TCG or NGSCB design.

Turkish), was born March 18. Andy and Ali Forsberg are now the proud parents of Mia, born June 3rd. And tstaffer Mark Dieterich and his wife Karen have a baby son, Nathan, born August 10. All are first babies!

------8<------

In a note to avd, David Salesin @ Microsoft (’83) tells of a visit by Ed Lazowska (’72) and

Susan Hutchison, who runs the Charles Simonyi Foundation, for a demo of ‘photomontage,’ a tool for recombining digital photographs. So lovely!

------8<------

changelog

After many years of network cable hodgepodge, Max and Jeff take advan-tage of a switch upgrade to organize the network closets. Images are before, during and after the job.

conduit! 12

They challenged the TCGand NGSCB advocates tocome up with example ap-plications that needed aTCG architecture featurethat was unattainable bysimpler means.

Another concern was that voiced by Seth,joined by many in the audience: the re-mote attestation feature of the TCG de-sign. The idea that an outside party candemand to know what program you arerunning and in what configuration seemeddangerous to them. This was not from thepoint of view of privacy. Indeed, our mainprivacy voice, Ari, expressed satisfactionwith Ernie’s and Jan’s (and HP’s LiqunChen’s) work in ensuring that the TCGtechnology did not compromise user pri-

vacy. Rather, the concern is that the tech-nology that lets a remote party find outwhat software you are running (eventhough ostensibly just to make sure youare running software that this remote partytrusts) is dangerous because it may resultin anti-competitive practices. Powerfulthird parties, such as banks, will end up re-quiring that their customers use certain ap-plications produced by specific vendors,thereby limiting competition.

I moderated the discussion, and so it fellon me to stop the debate to make sure thatthe yummy treats awaiting us in the Atri-um did not go unconsumed. But the dis-cussion continues offline, and I would becurious to receive readers’ opinions. I canbe reached at anna@cs.brown.edu.

Virus attacks and other Internet threatshave rapidly increased over the years andare causing major disruption to computerusers worldwide, with 2003 damage esti-mated at $55 billion, according to reportsby Computer Economics (CEI) and TrendMicro, Inc. But while 2003 was named theYear of the Worm, the number of computersecurity threats in 2004 will be much high-er. According to statistics compiled byMessageLabs, in February 2004 alonethere were 50 billion intercepted emailmessages containing virus code, the totalnumber of interceptions of 2003. Also, thelatest Symantec Internet Security ThreatReport states that in the second half of2003, five attacks originated from U.S.

computers for every 100computer users and that ex-port of confidential datafrom compromised ma-chines to the attacker (espe-cially passwords obtained bymonitoring keyboard activi-ty) is dramatically increasingas the payload of choice formalicious code.

The above gloomy statisticsreflect the multitude of ex-ploitable vulnerabilities inwidely deployed softwareapplications and operatingsystems, the general lack ofadequate knowledge about

security by both end users and softwaredevelopers, and the power of new attacktools produced by the recent alliance be-tween spammers and virus writers.

The silver lining behind the sad state ofcybersecurity today is that the business ofdefending computers and networksagainst threats is flourishing. Various sur-veys on workers’ compensation haveshown that computer security skills re-mained in high demand even during thetechnology downturn of 2001-2002 andcontinue to command large salaries today.In addition, investors in computer securi-ty c ompanies have been handsomely re-warded in the past five years. For

Danah Boyd (’01), currently a grad student at Berkeley, is making quite a name for herself. She was recently profiled in the NYT for her work with Friendster.com, whose millions of members have

transformed it from a dating site into a free-for-all of connectedness where new social rules are born of necessity. Danah studies Friendster with a Web log called ‘Connected Selves’ (www.zephoria.org/snt). Her observations have made her a social-network guru for the programmers and venture capitalists who swarm around Friendster and its competitors. Said

changelog

TEACHING COMPUTERSYSTEMS WITH A TASTE

OF DANGER

Roberto Tamassia Vesselin Arnaudov

The first in what we hope will be a series of articles by the faculty on the current curriculum.

conduit! 13

The course will be organized around acollection of projects and will use the fa-cilities of the Internet Computing Lab (de-scribed in the previous issue of conduit!).The course will complement our currentCS151, Introduction to Cryptography andComputer Security, which focuses on themathematical and computational founda-tions of security. In developing the newcourse, we are leveraging ongoing re-search collaborations with academic col-leagues at Brown and other institutionsand with our industrial partners. We hopethat our new course will serve as a blue-print for new computer-security coursesin other schools and will help give thenew generation of computer professionalsan increasing awareness and knowledgeof computer security.

Teaching a course on computer securityhas often proven to be both controversialand challenging. A first issue is the pre-requisites for such a course. Traditionally,computer security courses taught todayassume extensive computer science back-ground and require as prerequisites a vari-ety of junior/senior computer sciencecourses such as algorithms, operating sys-tems, computer networks, and softwareengineering. The typical assumption isthat students need an advanced knowl-edge of how computers and networksfunction and significant programmingability in order to start learning aboutcomputer security. This approach givesinstructors flexibility in selecting ad-vanced topics and projects. However, ithas led to small enrollments in computer

example, Symantec Corporation, a com-pany focused on security software and ser-vices, has seen its revenues grow from$650 million in FY1999 to $1.9 billion inFY2004 and its market capitalizationclimb from about $1.6 billion at the begin-ning of 1999 to over $13 billion today; itnow ranks among the top ten softwarecompanies worldwide.

A New Course

Motivated by the growing importance ofcybersecurity in today’s computer sys-tems, Roberto Tamassia and undergradu-ate research assistant Vesselin Arnaudovare developing a new undergraduatecourse, tentatively dubbed Introduction toComputer Systems Security, to be taughtin the 2005-2006 academic year. The newcourse will teach general principles ofcomputer security from an applied view-point, providing hands-on experience indealing with current security threats andavailable countermeasures. Students willlearn about common cyberattacks, includ-ing Trojans, viruses, worms, passwordcrackers, keystroke-loggers, denial of ser-vice, spoofing, and phishing. They willlearn how to identify and patch vulnerabil-ities in machines and networks as well asdetect and repair infected systems. Theywill study fundamental building blocks ofsecure systems such as encryption, finger-prints, digital signatures and basic crypto-graphic protocols. Finally, they will alsobe exposed to the human and social as-pects of computer security, including us-ability, interfaces, copyright and digitalrights management, social engineering,and ethical issues.

high-tech venture capitalist Joichi Ito, “She’s definitely a Pied Piper for a bunch of different people. At the same time she, as an academic, is able to articulate what is going on in a way that the people building the tools rarely understand or can articulate.” The CEO of tribe.net sought her advice because she is involved in some of the groups to which his site tries to appeal: “Danah’s this researcher, but she also lives the whole thing—the Burning Man scene, the rave scene, the techno music scene.” Her academic supervisors at Berkeley are envious of her advantage. Said Peter Lyman, “I look at cyberspace the way a deep-sea diver looks at the sea: through a glass plate. She is out there swimming in it.”

------8<------

changelog

Goals and Prerequisites

0.5 1.8 3.36.1

12.217.1

13.2

20

55

0

10

20

30

40

50

60

$B 1995 1996 1997 1998 1999 2000 2001 2002 2003

Estimated Virus Damage

conduit! 14

security courses and to a consequent shortage ofcomputer-security experts, since most computerscience concentrators graduate today without everlearning about computer security.

We will make our security course widely accessi-ble by setting as its sole prerequisites one of theintroductory sequences (CS 15/16 or 17/18). Wewill cover computer security while at the sametime providing background on the foundations ofcomputing. Thus, the course will serve the dualpurpose of teaching computer security topics suchas access control, firewalls, and viruses and intro-ducing a variety of fundamental computer-scienceconcepts especially in operating systems, net-working, and programming languages. We be-lieve it is possible to convey fundamentalcomputer security concepts and give students aworking knowledge of security threats and coun-termeasures by providing just-enough and just-in-time background CS material for their under-standing. The course will leverage and exercisethe student’s knowledge of programming and al-gorithms in the setting of information security. In-deed, both a solid programming discipline andefficient algorithms are essential for developingeffective security solutions.

Another reason to make our security course wide-ly accessible is to encourage our students to thinkabout security issues and deploy security mecha-nisms early in designing a software application.This skill will certainly be appreciated by their fu-ture employers, who include leading corporationsin the financial, health-care and technology sec-tors for whom the security of software applica-tions is often a critical requirement. Besidestraining information technology professionals insecurity, our course aims to create security-savvycomputer users who will have a clear understand-ing of the security ramifications of using comput-ers and the Internet in their daily life (e.g., foronline banking and shopping). Last but hardlyleast, motivated by the recent debate on electronicvoting, we want to make our students aware of thepotential threats to individual privacy, and possi-bly to our whole political process, that may arisefrom inappropriate computer security technology.

Special Challenges Two special challenges in teaching computer se-curity are the need to provide an isolated yet real-istic computing environment for safely experi-menting with security threats and defenses, andthe ethical and legal issues associated with teach-ing potentially hazardous knowledge to students.To address the first challenge, we will leverage thepowerful computing and network resources in ourInternet Lab. We will give each student a virtualcomputer network consisting of multiple virtualmachines with selected unpatched OS images thatwill be separated from the rest of the departmentalnetwork and will be completely under the stu-

dent’s control. Thus students will be able to designand inject threats into the network as well as de-ploy into the network tools for prevention, intru-sion detection and repair. Students can experimentfreely with and witness the devastating effects ofa cyberattack without having to worry about cor-rective action by our technical staff. Also, shouldthe student be unable to contain the attack orshould the student-crafted repair tool actually doeven more damage than the threat itself, a conve-nient “restart” feature will bring the virtual net-work back to working order.

Through a specially crafted course missive andmandatory student agreement and with extensiveTA training, we will also discourage illegal use ofthe knowledge and experience gained in thecourse.

Assignments An integral component of courses in our depart-ment is their projects. Besides making studentsspend many hours in a flux of alternating joy andfrustration, projects give a course its identity: theyprovide a concrete grasp of the subject matter andlinger in the students’ memories for a long time af-ter the course ends. We discuss below three sam-ple projects currently being developed for our newcourse.

The project code-named Sentinel will test stu-dents’ knowledge about autonomously propagat-ing system infections such as viruses and worms.Faced with a hidden yet mighty foe, an “educa-tional worm” developed specifically for thiscourse, students will perform forensic analysis ona snapshot of a live compromised machine, deter-mine the worm’s propagation, execution, and pay-load, and come up with an automated tool for its

removal. Our educational Java-based worm LittleKraken (named after the mythical Scandinaviansea monster) is fully configurable to spread auto-nomously in a controlled manner and deploy itspayload by launching various types of attacks on

conduit! 15

several operating systems. The students will beable to study its means of infection and reproduc-tion and be frustrated by and convinced of theharmful effects and dangers such infections pose.The project itself introduces students to operatingsystems, file systems, access control, fingerprints,execution environments and (indirectly) the dan-gers of buffer overflows.

Another project, code-named SpiderNet, willchallenge students to defend against a network-based denial-of-service attack. Traditionally, suchthreats are difficult to counter and require exten-sive knowledge of the network stack, protocolsand particular OS implementations. To make sucha project possible here, we have abstracted thelow-level handling of TCP/IP packets by the oper-ating system into a Java-based firewall frame-work. Thus students can focus on the efficientdesign and implementation of packet-filteringrules without having to understand the details ofhow packets are retrieved from the wire and han-

dled within the operating system kernel. Using thedesign of a simple firewall as a motivating chal-lenge, SpiderNet will introduce the important sub-ject of network architectures and protocols.

Our third project, Agent Bond, on the other hand,will introduce students to digital signatures andservice negotiation in the process of building anelectronic wallet. It will touch on the theory ofpseudo-random-number generators and publicand private key infrastructure, as well as the use ofapplication sandboxing. We plan to use JavaCard-compliant smart card devices to provide the hard-ware platform for this task.

In addition to projects, the course will have in-class demonstrations of commercial software ap-plications, focusing on their vulnerabilities andways of overcoming them. In addition, we areconsidering a possible field trip (!) on which wewould perform a so-called “war-drive” to scanand map unprotected available wireless networksin a city’s neighborhood.

JANET INCERPI BERTOT PhD ’86Hi Trina, I’ve just received in my mail (of the snailkind) a brochure for the upcoming celebration for25 yrs of CS at Brown. I’m not sure you will re-member me, but hell I’ll give it a try (you certainlyhaven’t changed in all the conduit! photos I’veseen over the years). I started in the Ph.D programin ’79, the dept’s first year and even got out(alive!) defending my thesis (working w/BobSedgewick) in August ’85. Then I slipped out ofthe country for a one-year visiting professor posi-tion at INRIA in France, where I should have spentsix months in the Antibes area and six months inthe Versailles area at two of the INRIA researchcenters. Well, that was 19 years ago and I still ha-ven’t made it up to do my six months in the Ver-sailles area! I remember Andy telling me, if I tookthis job, “they’d never be able to keep me down onthe farm” and he was right; so I stayed!

L’INRIA (Institut National de Recherche en Infor-matique et Automatique) is the French national in-stitute for computer science and control theory.I’ve been working as a research engineer (in soft-ware development, mostly in the domain of soft-ware and proof development environments) forvarious research teams ever since. I head a smallgroup of research engineers that work with the re-searchers doing software development and exper-imentation on various hardware platforms (robots,virtual reality, clusters) at the Sophia center.

I recently saw John Hughes who stopped in Sophiafor a final review of a European-funded graphicsproject. He said that you or Tom had mentioned Iwas in this corner of the world, which is true. Myadopted neck of the woods, as it were.

My email address is deceptive as I’m certainlylisted as Ms. Incerpi chez Brown! But my con-duit! copy always finds me... and as my email im-plies, there is a Mr. Bertot and even two little—well, they’re no longer that little, they’re tallerthan me—Bertot boys. I won’t be in the US inMay, but seeing the Alumni/ae Life Histories sec-tion of the program with Barbara, Norm, and Dil-ip, I thought I’d drop a line to say “hey there”.Pass on my hello to Tom, Andy, and the others. cheers (et bonne fête à tous pour le 27 mai!)

janet (Janet.Bertot@sophia.inria.fr)

RANDY CALISTRI-YEH PhD ’90Randy’s email is included in Eugene’s “CharniakUnplugged” column at the end of this newsletter...

RUSS ELLSWORTH AB ’79, ScM’85 Hello! Russ Ellsworth here—still working at Ray-theon in Portsmouth RI. I recently received “bestpaper” honors at the 2004 Raytheon Joint Sys-tems Engineering/Software Engineering Sympo-sium held in Los Angeles last March. The paper,entitled “Resolving Intermittent Failures: A Dis-ciplined Approach,” was co-authored with RobRaposo, a Raytheon systems engineer.

On the home front, my daughter Katrina is stillnamed after Trina Avery and she attends theAmerican Musical and Dramatic Academy onBroadway in NYC, which is as far from computerscience as she could get without actually leavingthe east coast; son Craig will be a second-genera-tion CS major, starting at Maris College in thefall. Wife Darleen is still a saint for having to putup with all of us...

ALUMNI/AE email

conduit! 16

STUART KARON, ScB ’85, ScM ’86The following email exchange ensued after Stuart Karoncontacted Suzi Howe to get in touch with pilot MarkStern, ’91.

Hello. I’m a Brown alum, computer-engineeringScB in ’85 and a CS master’s in ’86. In the latestissue of conduit!, you had an article about MarkStern. I was wondering if you could forward hisemail address to me or this message to him.

I got my private-pilot’s license a couple years agowhile living on the Navajo reservation in Arizona.I had a ’72 Cessna 172 that stayed in AZ when mywife and I moved back east to New Hampshire.We live a couple towns away from Mark. Due tohaving a child and the typical time one sinks intowork, I’ve yet to get behind the yoke of a smallplane since moving back east. I’d be interested inchatting with Mark about his flying experience inthis part of the country. I’d also like to offer myservices if he ever wanted company (and someoneto share expenses) while tooling around above theNew England hills.

Hoping Mark would respond, Suzi asked Stuart if he’dwrite an alum letter about their conversation—herquestions are in italics.

If it ever happens, I’d be happy to write somethingfor you. When I’m done developing and support-

ing software and running the little business thatgoes with it, I hope to do some creative writing.It’s something I’ve always enjoyed. Writing soft-ware requires creativity but it’s strictly logical. Inprose, you can explore much more than logic andfunctionality.

...Are you flying again?

I never enjoyed flying as an activity in itself.Learning to take off, land and otherwise controlan airplane was a challenge which I did enjoy. Imastered it enough to get a license and fly with asufficient level of safety. I never logged enoughtime to become completely comfortable flyinginto places I’d never visited before, though, andthat took some of the fun out of it.

In northern Arizona, the small plane was a greatway to cover long distances. We were in the FourCorners region and the closest commercial airportwas 3.5 hours away by car. We once flew to Albu-querque to catch a flight back east. We flew to LasVegas to meet my in-laws who were there for afew days. That would have been 6.5 hours by car.

Flying is an expensive and time-consuming hob-by and unless you’re crossing a number of states,it doesn’t even save much time. Between gettingto the airport, planning your flight, the pre-flightcheck of the plane and more, you can often behalfway to your destination by car before youeven get off the ground.

Steve Reiss’s recently completed Internet Lab on the newly reno-vated third floor. The floor-to-ceil-ing white boards, both in the lab and outside in the new lounge area, are well used. The lab is for experimenting with Internet-scale applications and is used both for teaching and research

conduit! 17

I hope to get back to flying one day but that mightnot be for a few years when my son, Holden, is alittle older. My wife and I bought an old farmhousein New Hampshire and it came with 30 acres ofland. Beyond our trees and fields are thousands ofacres of undeveloped woodlands. There’s plentyto do on the ground.

The previous owners left us with lots of gardensand flower beds. They take time to maintain andwe’re not quite keeping up with it all. The houseeven came with a sugar shack for making maplesyrup. I made syrup the first year we were here butI probably spent the equivalent of four or five daysto produce one gallon of syrup. I’ll get back intotapping trees and boiling sap when Holden is oldenough to participate. He’ll turn two next monthand isn’t quite ready to toss wood into the evapo-rator.

...living on the Navajo Rez in AZ sounds fascinating—more grist for the conduit! mill?

When viewed through the filter of some creativewriting, I’m sure it could make for interestingreading. I’m not sure anyone in the department ormore than one or two graduates would rememberme. My biggest claim to fame might be that I end-ed up marrying Dina Goldin, one of your morelong-term grad-student fixtures. That didn’t lasttoo long, though, and it’s ancient history by now.

As an undergraduate, I defied advice from Andyand others and wrote a space-flying video game asmy final project for Andy’s class (don’t even re-member the number by now). He didn’t look verykindly on video games back then. Perhaps themulti-billion-dollar gaming industry has softenedhis view a bit by now. It was a fun project and Isaw people playing the game on the Sun worksta-tions over the next year or two until they finallywiped it off the system.

Stuart Karon. http://www.spiralsoftware.com

JAMES PIECHOTA ’02Hey, Andy, James Piechota, here. I realize I’vebeen remiss in my promise to “keep in touch” af-ter leaving Brown. Here’s a whirlwind attempt tomake up for a year and a half of silence:

Graduated.

Cut my hair. Got a girlfriend (cause and effect?).

Moved to Toronto to work on Maya at Alias.Learned the cool people call the city “T-dot”.

Learned to drink beer (ranked #2 to hockey as anational pastime).

Took hip-hop dance lessons. Tried to keep thespirit of CS15 alive by putting on a “choreo-graphed” dance show with some other guys at ourcompany Hallowe’en party. Made a fool of my-self, and got a lot of laughs. A success.

Worked hard. Saw interns come and go. Thoughta lot about the good ole days of CS15.

Took some breakdancing lessons. Decided adance show would be pathetic and awkward.Wisely opted against a repeat performance.

Saw Bill Buxton leave the company and hoped hestill made it down for CS123 every now and then.

Worked hard. Saw interns come and go. Thoughta lot about the good ole days of CS15.

Decided the spirit of CS15 needed some reviving,and put on an almost-full-Monty show (socks,smiley-face boxers, and undershirts) at our sec-ond company Hallowe’en Party. Made a biggerfool of myself, got a lot of laughs and way toomany photographs. A success. Worked hard. Sawinterns come and go. Thought a lot about the goodole days of CS15.

Decided to branch out from self-humiliation and

VIP visitors from Industrial Partner Sun Microsystems display a 25th AnniversaryT-shirt outside the CITbuilding. L to r: Emil Sarpa, Jud Cooley and Joerg Schwarz. The folks from California experienced the start of a New England snowstorm. Despite their smiles, and no thanks todigital delay, they werefrozen by the end of the shoot!

conduit! 18

tried some sportier activities. Learned that thereare only two types of sports up here: indoor andsnow. Snowboarding hurts (in a good way), andwall climbing’s wicked awesome. It’s not sea kay-aking, but I think you’d dig it.

So now I’m still working hard, starting to feel alittle older than the interns, and wondering howCS15 has been going these past two years (name-ly: who played Fabio?).

Word on the street is that CS enrollments atBrown, and in fact at universities across Canadaand the US, are way down, especially in the introclasses. It seems people are naming the dot.combubble burst and current outsourcing trends aslikely culprits. As far as outsourcing goes: defi-nitely a concern for many of the developers I’vespoken to up here. Especially since, by all reports,the talent and skill level of the developer pool inIndia is at least on par with what can be found inNorth America.

I’m thinking, and this is pure conjecture, that oneof the major reasons we haven’t seen even moreexpansion into India by major software develop-ment companies is that the majority of sales arestill in North America. As such they’re always go-ing to need product specialists and product man-agers here as an interface to the end users. And, atleast for a while, that communication channelfrom user to product manager to developer is go-ing to be much better accomplished face-to-facethan through whatever means current idea-sharingtechnology can enable.

Then again, staking one’s career on the slow evo-lution of communication technology is not thesafest bet. I was reading an article about BillGates’ recent university tour to talk up computerscience, and in it Professor Guttag1 from MIT saidsomething to the effect of “Computer science is agreat preparation for almost anything you want todo”. I guess I’ve been aware of this all along, butsomething clicked when he reiterated it. Perhapsmore and more software development positionswill be moving abroad, but that doesn’t meanthere’ll no longer be a need for people savvy incomputer science.

It got me thinking about Sarah Papp (’02) and Zil-ian Cheuk (’02), both super-competent softwareengineers and both interested in careers outside ofsoftware development. And then I thought of ourown documentation team here at Alias. For themost part their degrees are in English, but thecomplexity of modern text layout applications(e.g. FrameMaker) means they are constantlytweaking scripts and managing complex layoutschemas. Heck, one member of the team wrotemost of our new documentation web server, com-plete with thousands of lines of javascript. And, asI’ve recently been reminded, Duncan Brimsmead,a brilliant and creative programmer who has cre-ated some of the most artistically innovative fea-

tures in Maya, got his degree in the French hornfrom Juilliard.

Maybe that’s the future of the industry? Comput-ing will become so pervasive that it will be astough to be a software engineer without knowl-edge of another field as it will to be a non-soft-ware engineer without knowledge of computerscience.

I wish there were some way to convince new stu-dents that CS15 and CS16 are not only for soft-ware engineers, but that they teach thefundamentals of a science that is increasinglywheedling its way into every profession.Hope all is well, and, as always, tx, James.

1. John Guttag ’71 recently gave the third of our four 25th Anniversary Distin-guished Lectures. See next article.

DAVID VORBRICH ’95Immediately after leaving Brown I took a road tripwith Shuang Ji (MS ’94) and Katuya Tomioka (ofElectronic Book Technologies). I then went towork for Citicorp helping deploy internal globalapplications, predominantly in South Americaand Asia. Who knew Lotus Notes wouldn’t last?At least the travel was exciting. I then went towork at EF Education (think Louise Woodward),a small education, culture, and travel companyupon receiving a job tip from Vince Rubino (BA’92, MS ’94). I found myself managing a networkof Macintosh computers, which some might evenview as a step back from my prior Lotus Notes ex-perience. Personally, I love Macs.

For the last seven years I’ve been working at Ac-centure (formerly Andersen Consulting) as an En-terprise architect within their financial servicespractice. I’ve worked mostly in the northeast withsome projects in Silicon Valley (back when dotcom companies had money). In addition to Shua-ng and Vince, I have also kept in loose contactwith David Langworthy (PhD ’95) and StanislavMarkovic.

My wife Lisa and I started our family a few yearsago, and we live in Grafton, MA with our twosons Gregory (3 years) and Nicholas (4 months).Feel free to contact me at david@vorbrich.com ifyou’d like to catch up!

conduit! 19

mit” meeting on Monday, May 24 under theauspices of the department’s Industrial PartnersProgram (it celebrates its 15th anniversary thisyear), and we held a symposium and banquet onThursday, May 27 as a reunion for students, facul-ty, alumni/ae and friends. The lecture series andsymposium were organized by a committee Ichaired whose members included Michael Black,Eugene Charniak, Tom Doeppner, Philip Klein,and Don Stanford. Michael Black and I organizedthe summit. Our distinguished lecturers were:

John Crawford ’75, Sept. 18, 2003: 20 Years of Growth in Microprocessor Perfor-mance: A Look Back and Glimpse AheadDavid Salesin ’83, Nov. 20, 2003: Next Frontier in Graphics: Unleashing the Computer’s Potential for CommunicationJohn Guttag ’71, March 4, 2004: Sensor-Based Medical Decision SystemsRobert Schapire ’86, April 22, 2004: Modern Approaches to Machine Learning

The May 24 IPP summit was entitled “The Na-tional Research Landscape: Leveraging the Aca-demic/Industry Partnership.” We were honored tohave the participation of the Director of BusinessDevelopment for the Rhode Island Economic De-velopment Corporation, research vice presidentsfor Hewlett Packard, IBM, Microsoft, MitsubishiResearch Labs and Sun Microsystems, and thepresident of Atomic Ordered Materials LLC.Michael Black describes the summit in the follow-ing article.

A Commencement Forum was held on SaturdayMay 29 to celebrate our 25th anniversary. Thespeakers were Joe Pato ’81, Distinguished Tech-nologist at Hewlett Packard Research Labs, andPascal Van Hentenryck and Stan Zdonik of ourfaculty. Joe described methods of protecting com-puters from malicious software. The magnitude ofthe problem is reflected by the fact that thousandsof computers can be infected in seconds. Pascal,who specializes in combinatorial optimization,gave a fascinating account of the importance of

John Crawford ’75 David Salesin ’83 John Guttag ’71 Robert Schapire ’86

In the Beginning On July 1, 1979, after nearly 15 years of research,teaching and awarding of degrees in the Divisionsof Applied Math and Engineering led by co-

founders Andy van Dam, PeterWegner and me, the Departmentof Computer Science came intoexistence. Shortly before its for-mal creation seven departmentalcolleagues moved into a newlyrenovated building at the corner ofGeorge and Thayer Streets, laterdubbed kassar House. ComputerScience has come a long waysince 1979. We have hired a stellarfaculty, now 24 strong, educatedmany generations of very talentedstudents, occupied the top twofloors of the Thomas J.WatsonCenter for Information Technolo-gy (we’re currently expandinginto the third floor), and developedan enviable track record as an aca-demic department.

In 1979 the field of computer sci-ence was mostly inward directed: it was still de-veloping concepts, frameworks, tools, analysisand theory so as to understand, use and controlcomputers. While many hard computational prob-lems still require solutions, this department andcomputer science as a field have become moreoutward directed today. Many of us are now inter-ested in solving the hard computational problemsarising in other fields. We are dealing with appli-cations that didn’t exist twenty-five years ago. Wehave evolved dramatically and will continue to doso.

Anniversary EventsWe celebrated the 25th anniversary of the depart-ment in three ways. We organized a lecture seriesin which four distinguished alums spoke, we dis-cussed industry/academic cooperation at a “sum-

CS CELEBRATES ITS 25TH !

John Savage and President Ruth Simmons enjoy a laugh at the 25th Gala cocktail reception

conduit! 20

his topic. One of the examples he gave is its usein scheduling football games for the NFL. Standescribed the data management problems thatarise in the new world of ubiquitous computing.Important and challenging data collection andmanagement problems will arise in this new con-text.

Anniversary Symposium

On Thursday, May 27, 2004, a day-long sympo-sium was held on campus to celebrate the depart-ment’s 25th anniversary. It was a wonderfuloccasion for students, faculty, alumni/ae, andfriends to reconnect with one another and withthe department. About 300 were in attendance.Eli Upfal, the fifth and current chair of ComputerScience, opened the meeting. To introduce our vis-itors to the department, we invited six of our new-est faculty arrivals to give twenty-minutepresentations on their research during the morn-ing. The titles of their talks illustrate the extent towhich our newest faculty members are outward di-rected. Tom Dean, the fourth chair during whoseterm these six faculty members were hired, intro-duced them.

David Laidlaw—Scientific Discovery Through VisualizationAmy Greenwald—Internet AgentEconomicsUgur Cetintemel—Mobile PervasiveComputingShriram Krishnamurthi—Constructing Robust SoftwareAnna Lysyanskaya—Trustworthy SystemsMichael Black—A Neural Motor Prosthesis for Augmenting the Damaged Brain

The theme of David Laidlaw’s talk is illustratedby a Fred Brooks quote that he cited: “Hitchingour research to someone else’s driving problems,and solving those problems on the owners’ terms,leads us to richer computer science research.”David’s many dramatic videos and images dem-onstrated the power of good user interface designand well chosen visualization paradigms for sci-entific understanding. David and his group areworking with biologists, physicians and physiciststo help understand phenomena through visualiza-tion.

Amy Greenwald described her research on thedesign of internet agents in which she uses a com-bination of decision theory and game theory, illus-trating this work by an example of biddingpatterns in eBay auctions that work to the detri-ment of the person offering an item for sale. Shewent on to describe her leading role in the annualinternational trading agent competitions.

Ugur Cetintemel gave examples of large-scalemobile networks in commerce and in the militarythat will emerge in the near term and described thechallenges of managing vast amounts of data thatthese networks will produce. Interestingly, these

data-management problems, while new, have ele-ments in common with traditional database-man-agement problems.

Shriram Krishnamurthi motivated his topic bydescribing errors that occur every day when ac-cessing a popular web-based travel reservationsystem. Shockingly, he reported that such errorshave been found in accessing prominent commer-cial and government web sites as well. He arguedfor the power of state machines to model and ver-ify programs and described his group’s efforts tointegrate design and verification tools and pro-gramming languages to build robust software.

Anna Lysyanskaya opened her remarks by ob-serving that, although we have pretty good meth-ods of assuring accountability and privacy in theanalog world, it is much harder to provide both inthe digital world in which our identities are repre-sented by our secret keys. She described a varietyof problems that arise with electronic signing ofdocuments and protecting digital identities, in-cluding her work on anonymous electronic certi-fication, and discussed progress in buildingtrustworthy systems and the problems that re-main.

Michael Black described his work with Brown’sBrain Sciences Program to develop neural pros-theses in the motor cortex. Translating the outputsignals from motor neurons into meaningful ac-tions by an arm, for example, requires the skills ofpattern-recognition experts. He sees great poten-tial for computer scientists to help develop bio-logically embedded hybrid neural-computersystems to help those impaired by disease and in-jury.

Two afternoon sessions were devoted to talks byseven alumni/ae. Each had very interesting storiesto tell. Andy van Dam, our first chair, moderatedthe first of the panels at which Ed Lazowska ’72,Barbara Meier ’83, ’87 ScM, Norman Mey-rowitz ’81, and Dilip D’Souza ’84 ScM spoke.

Astaffers man the souvenir T-shirt table. L to r: Trina Avery, Kathy Kirman, Lori Agresti,Fran Palazzo and Genie deGouveia

conduit! 21

Ed Lazowska amused the audience with tales ofescapades by Andy and his students and humorousphotos of many kinds. He remembered fellow stu-dents now deceased and expressed his gratitudefor the positive lessons he learned while at Brown,and also proudly showed a photo of the large con-tingent of Brown graduates who are now on thefaculty of his department at the University ofWashington.

While Barb Meier impressed the audience withthe range and quality of the computer graphics de-sign projects in which she participated, includingdocumentaries, movies, and the cover of SIG-GRAPH Proceedings, they were spellbound by herdescription of her personal struggles in decidingbetween raising a family and pursuing her profes-sional career. She left us awestruck by the sensi-tive and caring way in which she resolved thisissue.

Norm Meyrowitz marveled at both the advancesthat have been seen in technology since he was astudent in the late ’70s and early ’80s and at howlittle has changed in the range of ideas to which hewas exposed then and is using today. At Macrome-dia, where he is the principal developer, he hasbrought into existence many products that reflectthese ideas. For example, he drew connections be-tween Brown’s early BALSA animation packageand Macromedia’s Flash and between FRESS andIntermedia, two hypertext document linking sys-tems, and Macromedia’s Dreamweaver/Contrib-ute. He also drew an analogy between the cookiecan used to collect coins left in Andy’s couch withthe top-secret pouch that he was carrying aroundand would be part of an announcement during thecocktail party.

Dilip D’Souza has become a writer of consider-able note on political, social and human-rights is-sues. He has authored two books, Branded by

Law: Looking at India’s Denotified Tribesand The Narmada Dammed: An Inquiry intothe Politics of Development. Under Britishoccupation a number of Indian tribes weredeclared criminal in the 19th century and“denotified” under the Indian constitution.Although their status has long since beennormalized, they continue to be stigmatizedby their history. Dilip told about the buildingof a dam and the first introduction of electric-ity in Billdong, a town of 300 homes, byyoung electrical engineers; for him, theirwork exemplified a quiet patriotism that hefinds very appealing.

The second afternoon session was moderated byyours truly, the second chair of the department andthe organizer of the symposium. The speakers inthis session were Gloria Satgunam ’03, DonaldStanford ’71, ’77 ScM and Michael Littman ’96PhD. One other speaker was unable to attend at thelast minute.

Gloria Satgunam, who graduated last year witha concentration in computer science, representedthe new graduates. She attended high school inPawtucket where she was identified as a talentedyoung person by Don Stanford and encouraged toapply to Brown. She is now working for GoldmanSachs. Throughout her time at Brown and whilein New York she has been involved in volunteerwork with young people, giving back some of theopportunity that was given to her.

Don Stanford majored in international relationsat Brown because as an amateur radio operator hethought he already knew enough about the techni-cal world. It took several years of post-graduatework and a year as a law student to convince Donthat his calling was in fact in computer science.After intensive on-the-job training he entered ourprofessional Master’s degree program, graduatingin 1977. In 1979 he joined GTech Holdings Corp.in Rhode Island as CTO (their seventh employee)and remained there until his retirement in 2002.“GTech operates the world’s most reliable, secureand high performance transaction systems for thepurpose of selling online lottery tickets;” it has a70% market share and $1.1B in annual sales. Onretirement Don became an adjunct faculty mem-ber in CS where he teaches CS2, our course withthe largest enrollment.

Michael Littman, our last alumni speaker, joinedthe department in 1992 as a PhD student while anemployee at Bellcore, a spin-off of Bell Labs,graduating in 1996 with a thesis called “Algo-rithms for Sequential Decision Making.” Michaelthen went to Duke where he spent a very success-ful four years producing a couple of PhD students,earning an NSF Career Award, and starting theCrossword Project, the first automatic puzzlesolver. The latter earned him great fame and abest-paper award at the most competitive AI con-ference. He is now on the faculty of Rutgers andis very active in robotics, evolutionary learning,puzzle solving and reinforcement learning.

After the alumni/ae presentations we were hon-ored to be joined by Provost Robert Zimmer. Bob,who joined Brown in 2002, welcomed the audi-

Gloria Satgunam ’03

Don Stanford ’71, ’77

Michael Littman ’96

Vibha Kamat, Sahir and Dilip D’Souza ’84

Lunch on the green outside Starr Auditorium

conduit! 22

ence by telling us of the high standing that the de-partment enjoys in the university and sharing withus his expectations for the leadership role of thedepartment.

The afternoon was concluded with a cocktail partyin the CIT lobby at which President Ruth Sim-mons joined us. We all greatly enjoyed the conver-sation and camaraderie. At the end of the party EliUpfal presented me with a plaque recognizing myefforts in organizing this event, after which Presi-dent Simmons reflected on the accomplishmentsof the department since its inception, mentioningamong other things the three founders of the de-partment, Andy van Dam, myself, and Peter Weg-

ner. After she concluded her remarks, NormMeyrowitz made the surprise announcement thathe and a few other computer science alums had or-ganized a four-week fund raising campaign thatraised more than $4.1 million to fund an “AndyChair.” Andy was very pleased by this recognitionof his many years as an in inspirational member ofthe computer science faculty.

A banquet for 275 followed in Sayles Hall, afterwhich Eugene Charniak regaled us with an amus-ing slide show that chronicled the history of com-puter science at Brown since 1965. Some of uslooked awfully young and dorky. Because I failed

to provide a photo taken in the 1980s, Eugeneshowed me masquerading in a Santa suit!

A strong department and a great education are theresult of many hands. First among these is the fac-ulty. Thus, it is important to recognize the facultymembers who are not mentioned above but whoplay key roles in the success of the department. Inorder of their arrival at Brown they are Steve Re-iss, John Hughes, Roberto Tamassia, FrancoPreparata, Maurice Herlihy, Thomas Hofmannand John Jannotti.

Chairman Eli Upfal presents John Savagewith an award in recognition of his catalyzing

all the 25th anniversary celebrations

Andy van Dam flanked by President Ruth Simmons and Norm Meyrowitz after the

“Andy Chair” was announced

President Simmons addressing the gathering in the CIT lobby during the cocktail hour

Alumni at the banquet in Sayles Hall. Eugene’s slide show can be seen in the background of the photo at left

conduit! 23

“It’s not necessary for companies to invest inbasic research to make money from it.” Henry W. Chesbrough

Corporate investment in R&D isdeclining and companies are in-creasingly looking to academia toprovide research innovations. Howwill this shift affect the academic-industry relationship, and howshould computer science depart-ments like Brown’s respond andadapt to these changes? To addressthese questions, Computer Sci-ence’s Industrial Partners Program(IPP) hosted a special one-day

“summit meeting” on “The National Re-search Landscape: Leveraging the Aca-demic/Industry Partnership.” The summitbrought together leaders from industry,

academia, and government to explore trends incorporate R&D and to debate emerging modelsfor capitalizing on innovation.

The changes taking place in corporate R&D mightbest be characterized by the move to what is called“open innovation”. As Henry Chesbrough pointsout, “not all the smart people work for us.” Thisobservation has led to the idea that companiesdon’t need to invest in basic research to benefit

from it. While this is obviously appealing with to-day’s focus on containing costs and increasingprofits, is it wise? Is it good for industry? Are uni-versities willing partners and beneficiaries? Thesummit speakers explored these and similar ques-tions.

Merck, like other companies, increasingly scoursjournals, conferences, and patents and then evalu-ates the developments, picks the best, and pursuesthem further. Procter and Gamble implementsopen innovation using technology “scouts” whosearch the world for good ideas. Dow, like BASFand others, is farming out its problems to theworld by putting them on the InnoCentive web-

site; now anyone can try to solve their open prob-lems for a cash prize. With such outsourcing ofinnovation, companies have downsized their com-mitment to in-house basic research. Xerox, for ex-ample, once had one of the nation’s premierresearch labs but has now spun it off into an inde-pendent company. Intel, which long shunned basicresearch, has instead reinvented it in lablets thatblur the boundary between corporate and academ-ic research. These lablets are associated with uni-versities, are run by academics on leave, and arestaffed with both graduate students and Intel em-ployees. Microsoft is bucking these recent trendsby investing heavily in traditional research labs;this approach may be accessible only to near-mo-nopolies with the deep pockets necessary for along-term view.

According to NSF, corporate spending on researchand development dropped from $198.5B in 2001to $190.8B in 2002 (4.9 percent in inflation-ad-justed dollars), the largest single-year declinesince the NSF started keeping track in 1953. Cor-porations did not, however, compensate for theirdeclining internal research investment by increas-ing academic research funding: this actually de-clined by 1.2 percent over the same period. Incontrast, federal funding for university-basedR&D increased 13.6 percent in 2002, the largest

such increase since 1979. These data suggestthat the burden of corporate R&D is movingto universities and the cost is being funded bytaxpayers. They also suggest the need for co-ordination between academic an industryleaders to mediate what are bound to be con-flicting goals.

Enabling the move to open innovation is thefact that universities have a responsibility tosee that government-funded research is prop-erly commercialized for the benefit of soci-ety. The Bayh-Dole Act gives universities anentrepreneurial mandate and encourages aca-demic-industry collaboration. Many worry,

however, that by focusing on commercialization,academia may lose some of the freedom thatmakes it an incubator of ideas. It may also changethe academic mission. A recent study published inthe January 2002 Journal of the American MedicalAssociation reports that 21% of geneticists with-hold information from other researchers to protectits commercial value. There are also dangers forindustry if, by relying on academia for innovation,they lose the ability to drive research to solve theirbusiness problems.

To focus the summit discussion, we posed a num-ber of questions to our speakers:

Is the industrial research lab disappearing?

INDUSTRIAL PARTNERS PROGRAM SUMMIT: THE AMERICAN RESEARCH LANDSCAPE

Enabling the move to open innovation is the fact that universities have a

responsibility to see that government-funded research is properly

commercialized for the benefit of society

Professor Michael Black, co-director of the Industrial Partners Program and co-

host of the IPP Summit

conduit! 24

Can industrial research be outsourced? What are the best ways to advance industrial re-

search over the next decade? Is curiosity-driven industrial research dead? What role should government play in funding

corporate research? Have industrial affiliates programs failed indus-

try? Can they, and should they, be saved?

Addressing these, and other, questions were thefollowing distinguished speakers:

John Seely Brown ’62, former head of XeroxPARC, former Chief Scientist of Xerox, andcurrently Visiting Scholar, Annenberg Center atUSC

Per-Kris Halvorsen, VP & Director, Solu-tions and Services, Research Center, HP Labs

Alfred Spector, VP Services and Software,IBM Research

Robert Sproull, Fellow and VP, Sun Micro-systems

Jack Breese, Director, Microsoft Research Richard Waters ’72, President and CEO,

MERL Saul Kaplan, Director, Business Develop-

ment, Rhode Island Economic DevelopmentCorporation

John Preston, Senior Lecturer, MIT

According to John Seely Brown, the key lessonof industrial research labs is that research must bedeeply rooted in real problems for it to have an im-pact. Innovation, Brown tells us, is easy; turning itinto a successful business is the hard part. The oldmodel of innovation “thrown over the wall” to de-velopers is clearly flawed. In his words, “knowl-

edge flows on therails of practice;”and communities ofpractice are neededto develop sharedbeliefs and trust be-tween innovatorsand developers. Infact, Brown pointsout the false dichot-omy between inno-vators and devel-opers—the processof developing a

product and the associated manufacturing process-es is full of innovation, possibly more than was in-volved in the initial idea. The university is notsomething special in terms of innovation; instead,we are all producers and consumers of knowledgeand innovation. From his experience as head ofXerox PARC, Brown likens the successful re-search laboratory to an “ivory basement”. In con-trast to the iconic and much maligned ivory tower,the ivory basement is rooted in practice, bold butgrounded, and fundamentally cross disciplinary.

Kris Halvorsen of HP shed more light on the ac-ademic funding picture. The increases we haveseen in federal funding during the last 30 years arealmost entirely in the life sciences, with only slightincreases for math and computer science. He alsopointed out that, while the US funding picture isbleak, other countries have been dramatically in-creasing investment in R&D. He sees this as partof a move from a multinational business model, inwhich you invent once then sell worldwide, to ameta-national model where business inventswhere the skills and needs converge. Asian coun-tries are producing more and more PhDs, andmore of them are staying in their home country orreturning after a PhD abroad; Chinese universitiesgranted 465,000 science and engineering degreesin 2001, approaching the U.S. total. While Hal-vorsen sees industries embracing the open-innova-tion view of the research university as a source ofgraduates and applied research, he sees a crisisemerging around how universities protect andmanage intellectual property (IP). His view, whichwas echoed by most of the industry speakers, isthat universities are too aggressive about gettingvalue out of their IP and thus discourage compa-nies from picking up on academic inventions. Hepoints out that few universities have realized sub-stantial returns on their IP. Since there have beenso few successes, the argument goes, it makessense for academia to forgo IP as a means of gen-erating revenue.

For Microsoft the story is somewhat different andmore traditional. Jack Breese describes a good-old-fashioned research laboratory in which re-search and development are tightly coupled.While maintaining a strong internal research ef-fort, Microsoft also reaches out to universities, of-ten in creative ways. The big problem for

Microsoft is finding welltrained computer scientiststo satisfy the needs of a mul-tinational R&D effort.Breese described the diffi-culty of hiring 100 softwareengineers at their Asia Ad-vanced Technology Center,Beijing, despite receiving120,000 applications fromall around China. In lookingat the industry/universitypartnership, Microsoft’s

During breaks between sessions

conduit! 25

focus today is on training and recruiting futureemployees, and a significant focus of its academicfunding involves innovations around education.Like other companies, when Microsoft looks touniversities for IP, it prefers a “non-exclusive,non-transferable, worldwide, royalty-free license”with an option for an exclusive license. In manycases there is a safer model for companies like Mi-

crosoft: to wait for en-trepreneurs to take therisk of developing newproducts and then buythe company once thekinks are worked out.

As with Microsoft,MERL’s Richard Wa-ters sees in-house basicresearch as essential fora large company such asMitsubishi. Research atlabs such as MERL canbe focused on Mitsub-

ishi’s specific needs andcan follow through onideas. Waters pointed out

that basic research in computer science is oftenmuch more applied or easily applicable than basicresearch in the physical sciences. Despite this, re-search with a long time horizon is difficult to sus-tain and a successful lab must maintain a balanceof work with some being directly applicable today.To make innovations successful requires a tightcoupling between researchers and developers(Brown’s ivory basement), and this makes IP de-veloped at universities less valuable than in-houseIP. Like Microsoft, MERL sees the power of uni-versities in the students we train, in faculty con-sulting relationships, and in the collaborations

between MERL researchers and their academicpeers. Most forms of direct funding to universitiesprove too expensive for MERL with too little pay-off—Waters has yet to see a university/industryconsortium worth its cost. If, like MERL, a re-search lab has top researchers in a field, then mem-bership in an industrial affiliates program is reallyunnecessary to gain access to university research-ers and students. Like Breese, Waters sees the ag-gressive stance that university licensingdepartments are taking with respect to IP as coun-terproductive. He feels that universities overvaluetheir IP and ask too much for it; as a result, theylimit collaboration.

Robert Sproull recognizes that good ideas maycome from outside Sun’s research labs but the labsfunction as the eyes and ears of the company. Theapproximately 100 people in Sun Labs provide animpedance match between academia and Sun bybeing both comfortable with basic research anddeeply immersed in Sun’s business. Consequentlythe focus of academic partnerships is on people—interns, visiting scientists, and collaborativeprojects.

At IBM, research labs provide a talent pool thatcan funnel information into IBM. But even atIBM, with its long-standing culture of valuing re-search, the demands and expectations are high.Alfred Spector pointed out that a company thesize of IBM is looking to create new $10 billionbusiness. This is a tall order that may be more eas-ily satisfied without research; for example, IBMsold its disk-drive business and used the proceedsto buy PricewaterhouseCoopers. Along similarlines, Spector sees IBM’s highly respected re-search as something that could be sold as a ser-vice—IBM scientists could provide consulting

IPP Summit speakers and hosts: l to r: John Savage, keynote speaker John Seeley Brown, Richard Waters, Saul Kaplan, John Preston, Robert Sproull,

Michael Black, Alfred Spector, Jack Breese, Per-Kristian Halvorsen

Speaker Jack Breese and Deputy Provost Tom Dean share a laugh over dessert

conduit! 26

services either in house or exter-nally for a fee.

Saul Kaplan of the Rhode IslandEconomic Development Corpora-tion (RIEDC) discussed the role ofgovernment in fostering innova-tion. As RIEDC’s Director of Busi-ness Development, he provided aview of how a small state such asRhode Island has unique qualitiesthat let businesses “think big, startsmall, and scale fast.” He arguedthat RI’s compact geography, di-verse population, and accessiblegovernment allow companies totry out ideas here before invest-ing nationally. As an example hedes-cribed the development of hy-drogen-powered cars and the diffi-cult task of building the necessaryinfrastructure for delivering hydro-gen to consumers. Here RI’s smallsize is an asset: with only 350 gasstations, the environment is appropriate to devel-oping and evaluating the necessary technologyand infrastructure. Kaplan also argued that gov-ernment should be seen as an active partner in theindustrial/academic relationship. In particular, thesmall size of RI’s government makes it accessible,flexible and creative.

John Preston focused on how universities canspin off companies, a subject he knows well, hav-ing crafted MIT’s patent and licensing policy. MIThas spun off over 4000 companies worldwide, in-cluding 1065 in Massachusetts. In 1997 thesecompanies had over $237 billion in revenues,making them equivalent to the 27th largest econo-my in the world. He pointed out that 70% of newjobs in the US come from 4% of the companiesand that universities have a role in spawning these“gazelles”. In contrasting European universitieswith their U.S. counterparts, Preston arguedstrongly that U.S. universities are well positionedto spin off technology, engage new businesses,generate capital, and create entrepreneurs. Likemany of the speakers, Preston worries that univer-sities interested in short-term revenues are stiflingnew businesses and discouraging investment in IP.He argued that universities should prefer equity innew companies over licensing revenues, so as toown a small part of a potentially large pie. In lis-tening to Preston, one could argue that universitiesshould focus less on partnering with companiesand more on partnering with venture capitalists.Bold innovations are the lifeblood of entrepre-neurs while they may fall on deaf ears at a largecorporation. This also suggests a possible role foruniversities in helping larger corporations connectwith venture capitalists in the context of new tech-nologies.

At the end of the day, the emerging consensusamong the business leaders was that the primary

Eugene Charniak (l) holds forth with program speakers and John Schuster (r) from CS Industrial Partner Network Appliance. The sculpture in the background is a Lichtenstein called “Brush-

strokes.” It is on loan to the University for two years

benefit universities provide industry is in trainingnew employees. The value of access to university-generated IP was less clear. As Halvorsen pointedout, large technology companies increasingly fo-cus on shifting from products to services, and ser-vices do not lend themselves to improvementsfrom R&D the way products do. He also noted thatthere are no examples of great research environ-ments in small or mid-sized companies and thatthis may provide an opportunity for universities.

Both industry and academic (Preston) speakersagreed that university licensing policies were tooaggressive and needed to be changed to spur tech-nology transfer. It seems that industry would pre-fer not to pay for innovation; their ideal modelappears to be one in which the government fundsacademic research that is then given freely to in-dustry. The idea has undeniable appeal to industrybut remains fundamentally flawed as a model forsustainable innovation. While industry may be re-lying more and more on academia for innovation,our IPP summit meeting revealed that the modelunder which this work is funded, innovation is re-warded, and business people and academics col-laborate remains poorly understood. One of thegoals of IPP over the next fewyears will be to work with ourpartner companies to formu-late a viable model that worksfor both parties. The summitwas a first, and very illuminat-ing, step in that direction.

conduit! 27

Professor Paris Kanellakis (1953-1995) was amember of the CS department from 1981 until histragic and untimely death in an airplane crash in1995, along with his wife Maria-Teresa and twochildren, Alexandra and Stephanos.

2003 was the year of Paris’s50th birthday. To commemoratehis legacy to computer science,a memorial workshop was heldin San Diego on June 8. Theworkshop, formally called‘Principles of Computing &Knowledge: Paris C. KanellakisMemorial Workshop on the oc-casion of his 50th Birthday,’and abbreviated to ‘PCK50’,was a retrospective of his workand a celebration of his impacton computer science throughhis research and its influence onresearch directions taken by thecomputer science community.

The workshop consisted oftalks by invited speakers and byseveral of Paris’s past students.

The invited speakers included Paris’s advisor, Dr.Christos Papadimitriou, and his close colleague,Dr. Moshe Vardi. The highlight of the workshopwas a talk by the winner of the ACM 2001 ParisKanellakis Theory and Practice Award, Dr. GeneMyers of the University of California at Berkeley.The day was capped by a banquet at which col-leagues, former students, and friends offered theirpersonal recollections of Paris.

PCK50 was organized by three of Paris’s PhD stu-dents, Alexander Shvartsman, Dina Goldin andScott Smolka, and two of his close colleagues, JeffVitter (Purdue) and Stan Zdonik (Brown). The

workshop was affiliated with the ACM FederatedComputing Research Conference (FCRC).

It was sponsored by the ACM Special InterestGroups (SIG) on Management of Data (SIG-MOD) and Algorithms and Computation Theory(SIGACT). The workshop was generously sup-ported by Paris’s parents, Eleftherios and RoulaKanellakis, the University of Connecticut, and theComputer Science Department at Brown.

The proceedings of the workshop were publishedby ACM and included an In Memoriam articlewritten by Drs. Serge Abiteboul, Gabriel Kuper,Harry Mairson, Alex Shvartsman, and Moshe Var-di; the abstracts of the invited talks; and the paperspresented at the workshop.

Said Todd Millstein ’96, “It was a fabulous cele-bration of the life and legacy of Paris Kanellakis.”The workshop began with an invited talk byChristos Papadimitriou on “the new problems,”the application of several areas of theoretical com-puter science to the study of the Internet. Pa-padimitriou conjectured that Paris would haveloved these problems, which combine several ofhis interests and areas of expertise, and that hewould doubtless have been a leader in this field.

The remainder of the workshop consisted of talksby former colleagues and Paris’s students, dis-cussing research inspired by their work with him.Paris’s incredible range was evident, with topicsincluding logic databases, constraint databases,parallel and distributed algorithms, and computa-tional complexity. As in Papadimitriou’s talk, itwas amazing how often speakers intimated thatwere he here, “Paris would no doubt be workingon these kinds of problems”—a testament to hisbroad reach and impact across computer science.”

PCK50—A TRIBUTE TO PARIS KANELLAKIS’ 50th BIRTHDAY

A composite image of the ACM conference. l to r: Alex Schwartzman, Chryssis Georgiou, Iris Bahar, Peter Wegner, Dina Goldin, Peter Revesz and Gosta Grahne. Credit and thanks for the image to Peter Revesz

conduit! 28

MICHAEL BLACK. Michael was promoted to full professor effectiveJuly 1. He also joined the Editorial Board of theInternational Journal of Computer Vision and he

and Ben Kimia (Engineering) ed-ited a special issue of the journalfocused on computer vision re-search at Brown.

Conference travel took Michaelto Nice, Cancun, Vancouver,Whistler, and New Orleans. Be-fore the conference in Nice,Michael and his wife managed tosqueeze in a weekend in HauteProvence. They stayed at La Bas-tide de Moustier, a small countryinn run by Alain Ducasse, wherethe food and hiking are superb.At Whistler, Michael gave an in-vited talk in a workshop on ‘OpenChallenges in Cognitive Vision.’Also he gave a keynote talk at theInternational Conference on Ma-chine Learning in Banff in Julyon ‘Learning to See People.’

Michael and his collaborators in Neuroscienceand Engineering received new funding for theirwork on neural prostheses from the Office of Na-val Research and the National Institutes of Health.They also received funding as part of a large teamof European researchers working on neuroboticswith the goal of combining ideas from roboticsand neuroscience to improve human health.

Renovations are now complete on Brown’s videoand motion-capture facility. This unique facility ison the first floor of the CIT. It houses equipmentfor the analysis of human motion used for re-search on vision, graphics, and learning, and hasenabled Michael and his students to develop newalgorithms for detecting and tracking people invideo streams.

TOM DEAN.Tom’s new book Talking with Computers is nowon the shelves. In it, he explores a wide range offundamental topics in CS, from digital logic andmachine language to AI and the Web. He contin-ues his duties as Deputy Provost and still managesto teach a freshman seminar.

AMY GREENWALD.This summer Amy attended two conferences inBanff and followed them up with a week of hikingand camping in Jasper National Park with her hus-band and one-year-old daughter, Ella. (It was El-la’s second time sleeping in the great outdoors.)She gave a tutorial on Game-Theoretic Learningat ICML and presented a paper on Bidding UnderUncertainty at UAI.

Amy headed straight from Canada to NYC, whereshe and her TAC posse (Victor Naroditskiy,Jonathan Bankard, Bryan Guillemette, Haru Sakai(UTRA), and Lucia Ballard (UTRA)) attended thefifth annual Trading Agent Competition at AA-MAS. Botticelli, Brown’s TAC SCM entrant,placed first in its semi-final heat. RoxyBot,Brown’s new and improved TAC Classic entrant,was a finalist once again.

SHRIRAM KRISHNAMURTHI. Shriram has now served on the PCs of both FOALand FOOL. He looks forward to invitations from

fac.activities@cs.brown.edu

Michael Black (center) shares the deservedly underappreciated Lucia Prize for best talk at the NIPS2003 Workshop on “Open Challenges in Cognitive Vision”. Co-winners l to r: Andrew

Zisserman, Oxford; Jitendra Malik, Berkeley

Haute Provence

conduit! 29

FOIL and FOUL, which surely also exist. And ifall else fails, he plans to create a workshop on theFoundations of Ornamentally Extravagant Lan-guages. (Getting submissions shouldn’t be hard.).

Shriram served on the usual pile of program com-mittees, including PLAN-X, AOSD, ECOOP andPASTE. The AOSD PC meeting took him only asfar as Boston, but ECOOP lured him to Switzer-land. While the PC meeting and a research visit toBern were fun, they didn’t compare as attractionsto the Alps. He was fortunate to have cool, crispand sunny days for most of his time. He spent sev-eral hours just gazing at the Bernese Oberlandfrom the Uetliberg Hill, outside Zürich, and thengot to take them in while sweeping down the westof the country, before switching to the FrenchAlps in Lausanne. Lausanne also holds the Col-lection de l’Art Brut, a refreshing kind of artmuseum.

DAVID LAIDLAW. Congratulations to David, who received tenurethis year. He was awarded another ITR from NSFlast October, and is gearing up to teach his jointBrown/RISD class, Interdisciplinary ScientificVisualization, again this fall. He gave talks at Im-aging in 2020 at Jackson Hole, the Winter Confer-

ence on Brain Research,Copper Mountain, CO, aDagstuhl seminar in Ger-many, and several other

venues. Together with Sharon Swartz (Ecology &Evolutionary Biology) and Kenny Breuer (Engi-neering), David has received a Salomon Awardfrom Brown for his interdisciplinary work on theaerodynamic mechanisms of bat flight. This mul-tidisciplinary collaboration holds promise foridentifying principles that might help in creatinganimal-sized flying machines.

ANNA LYSYANSKAYA. Anna has started advising Math/CS concentrators.The result is they have told her everything aboutthe department and she now finally feels that sheknows the ropes fairly well—perhaps the infor-mation is flowing in the wrong direction! Thisyear, she will also be serving as a CAP advisor,which she’s really looking forward to. Her otherrecent activities include serving on the Eurocrypt2004 program committee, hosting the spring IPPSymposium, teaching CS22 (says Anna, “My TAsrock!”), and coordinating the Theory Colloquium.To ice the cake, she has also been awarded an NSFCAREER grant.

With support from the CS Department, eleven Brown studentmembers of the National Society of Black Engineers (NSBE) at-tended the NSBE’s annual conference earlier this year, includingten engineering majors and one CS major. The mission of the NSBEis to increase the number of culturally responsible black engineerswho excel academically, succeed professionally and affect thecommunity positively. As NSBE’s key event, the national conventiondraws thousands of black engineers from schools across the coun-try to network, find internships and jobs and work on resume-writ-ing and interviewing skills.

This year’s conference fair featured over 400 hi-tech corporations,government agencies and non-profit organizations all looking forpossible hires. While the engineers had mixed feelings about thejob fair, it proved extremely helpful for the CS major. Having a goodGPA and studying in the Brown CS department made him a prime

candidate. By the time the dust had settled a week later, he hadhad six interviews and four summer job offers!

Thanks to Chipalo Street, Brown’s NSBE CS Consultant and Kelly Jackson, Brown’s NSBE President, for this sidebar.

conduit! 30

MAURICE HERLIHY. The 2004 Gödel Prize for outstanding journal ar-ticles in theoretical computer science was sharedby Maurice for his paper “The Topological Struc-ture of Asynchronous Computability” coauthoredwith Nir Shavit, and by Michael Saks and FotiosZaharoglo for their paper,“Wait-Free k-Set Agree-ment Is Impossible: The Topology of PublicKnowledge.” According to the citation, “Thesetwo papers offer one of the most important break-throughs in the theory of distributed computing.The problem attacked is the complete understand-ing of asynchronous wait-free deterministic com-putation in the basic shared-memory model.These papers demonstrate that one can avoid theinherent difficulty of analyzing a dynamic model,transforming it into a static one by associatingcomputational tasks with simplicial complexesand translating the question of existence of a wait-free protocol into (distinct but related) topologicalquestions about the complexes. This reformula-tion allows the introduction of powerful topologi-cal invariants, such as homologies, to show theimpossibility of numerous tasks, including set-agreement and renaming. The discovery of the to-pological nature of distributed computing pro-vides a new perspective on the area and representsone of the most striking examples, possibly in allof applied mathematics, of the use of topologicalstructures to quantify natural computational phe-nomena.”

Next year Maurice and his family will spend asabbatical in England at Cambridge. His childrenwill attend school locally and aren’t too keen onthe idea of wearing school uniform!

JOHN SAVAGE. During the last academic year, as reported above,John chaired the department’s 25th anniversarycommittee, which organized not only the gala onMay 27 but also our anniversary distinguishedlecture series in which four outstanding alumnispoke. Michael Black and John, co-directors ofIPP, organized the May 24 IPP summit thatMichael summarizes in this issue. John alsochaired the search committee for Vice Presidentfor Public Affairs and University Relations, acommittee whose members included four mem-bers of the Brown Corporation, four faculty mem-bers, and six senior administrators. Thecommittee successfully completed its work whenthe president appointed Michael Chapman, for-merly of NYU Medical Center to fill the position.

On July 1 John started a sabbatical leave that he’llspend in part in Paris at L’Ecole Polytechnique.The year will be devoted to research on computa-tional nanotechnologies. He has just received a$1.3 M four-year NSF Nanotechnology Interdis-ciplinary Research Team grant that funds mywork and that of a chemist at Harvard and a com-puter scientist/electrical engineer at Caltech.

ELI UPFAL. Together with Adam Kirsch (ScB ’03, now a grad-uate student at Harvard) and Aris Anagnostopou-los (his graduate students), Eli gave a paper at the44th Annual Symposium on the Foundation ofComputer Science (FOCS ’03) in Boston. He willbe the chair of the program committee for thisconference next year, in Rome. Eli visited his ex-graduate student Gopal Pandurangan, now an as-sistant professor at Purdue, and gave a talk—hisfirst visit to the “real” midwest!

ANDY VAN DAM. Andy gave a talk at Brown about visualization ac-tivities in the CCV (Center for Computation andVisualization), then TCASCV (Technology Cen-ter for Advanced Scientific Computing and Visu-alization), entitled “Visualization: New Dimen-sions, New Domains, Old Questions”. This springhe gave a keynote, “Distributed Computing forGraphics: Then and Now” in Japan at the IEEEICDCS conference, which he co-founded as aworkshop at Brown in 1985 (his co-founder washis then Ph.D. student Jack Stankovic, now guruin the field and department chair at UVA). In be-tween, in addition to his teaching and VP Re-search responsibilities, he managed to spend aweek backpacking in the Grand Canyon and aweek scuba diving with his family (includinggrandchildren) in Bonaire, and managed to get insome good dives.

conduit! 31

Here is another in our seriesof articles by the faculty on thecurrent curriculum.“How long do you think it’ll takea pair of students?”, I askedUgur. “Oh, about a week,” he as-serted. Perfect, I thought. I hadjust shy of 20 students. One weekfor two students meant about tenweeks for 20 students. Just thelength I was looking for.

Brown CS has two courses that would looselybe classified as “software engineering” inother computer science departments. One is asophomore-level course, CS032, that intro-duces students to medium-scale program de-sign, group programming, and a collection ofconcepts (depending on instructor) rangingfrom network programming to low-level

memory management. The other course,CS190, is meant to be a senior capstonecourse where students learn to work in largergroups, build more significant products, gath-er requirements, and so on. This course reportis about the latter course as taught last springsemester. (Brown has one more course, RogerBlumberg’s CS092, on building educationalsoftware. It’s a real software engineeringclass in its own right, but really is unique toBrown, and therefore not one you might findelsewhere.)

In recent years, CS190 has been taken annu-ally by 20-30 students.Initially, the students work in small groups topresent ideas of projects they’d like to build.The class votes on these projects to identifythe most popular ones. The students then splitinto groups of 8-10 students each, and each

COURSE REPORT, CS 190:WE KNOW WHERE YOU ARE!

ShriramKrishnamurthi and menagerie

CS224 STUDENTS WIN @ SIGGRAPH ! CS224 (Spring ’04) student papers won first prize, second prize, and one of the two under-grad research awards at the ACM SIGGRAPH ’04 Student Research Competition held inLA in early August. Two more CS224 student papers made it to the semifinals (25 semifi-nalists out of 118 accepted submissions).Morgan McGuire (G), Andi Fein (’04) and Colin Hartnett (’04) won first prize and a total of

$750 for their paper “Real-Time Cartoon Rendering ofSmoke”. Pawel Wrotek (’05), Alexander Rice (’05), and Mor-gan McGuire (G) won second prize and a total of $550 fortheir paper “Real-Time Bump Map Deformations.” GabrielTaubman (’05) and Edwin Chang (’05) won one of the twoundergrad research awards and $500 for their paper “A FastFracture Method for Exploding Structures.” The following grad students were awarded $250 per paperas semifinalists: Peter Sibley, Philip Montgomery and LizMarai for their paper “Wang Cubes for Video Synthesis andGeometry Placement”; and Ethan Bromberg-Martin (ugrad’05), Arni Jonsson, Liz Marai and Morgan McGuire for theirpaper “Hybrid Billboard Clouds for Model Simplification”.Congratulations are definitely in order to the grad TAs whotaught CS224 under Andy van Dam’s direction: Tomer Mos-covich, Liz Marai and Morgan McGuire—well done indeed!Winners of ACM student research competitions held at ACMSpecial Interest Group conferences throughout the year(SIGGRAPH, SIGPLAN, SIGOPS and SIGCSE) will competeagainst each other in the ACM Grand Finals in early 2005.The winners of the Grand Finals will be recognized at theprestigious annual ACM Awards banquet.

Immediately before the conference Morgan McGuire also won a $25,000 fellowship fromNVIDIA Corporation for his research on hardware accelerated graphics.

L to r: Pawel Wrotek, Liz Marai, Gabe Taubman,Andi Fein (front), Tomer Moscovich (back), Edwin

Chang, Morgan McGuire,

conduit! 32

group takes on one project. The projectsrange from games to Web services to soft-ware development tools.I changed two things in CS190 this year.First, I wanted the students to deal with prob-lems in integrating large components, espe-cially ones that were themselves evolving.The best way to tackle this was to have thegroups work on different parts of the samesystem, thereby (inadvertently) creating trou-ble for one another. The second was to choosea project for them, to better reflect practice.Of course, I then needed to find a project thatwould be interesting enough to hold their at-tention (and get over their disappointment atnot being able to push their own favorite con-cept).That’s where I got lucky. In mid-fall, AbigailRider contacted the department with a pro-posal for a course project. Abbi is a directorat Brown whose portfolio includes Brown’srecently overhauled SafeRIDE service.SafeRIDE is a motorized escort with twoparts. One is a fleet of escort vans that runs ona fixed route every fifteen minutes or so. Theshuttles, in contrast, respond to calls from rid-ers who need to travel between specific loca-tions on- and off-campus. There are currentlyfour shuttles, and the SafeRIDE dispatchermust group together calls to plan a route foreach shuttle, evolving the route as new callsarrive. Abbi noticed that the dispatcherssometimes lose track of where a vehicle is,necessitating repeated calls to find out its lo-cation. She felt it would be a lot easier if thedispatchers could instead track the location ofeach van using some sort of map-display pro-gram. She referred the problem to us.It was immediately clear that this problemwas perfect for our needs.

First, it involved real users: students wouldhave to interview actual clients, most ofwhom would be nontechnical (drivers, dis-patchers, riders, supervisors, ...) and then de-ploy their product to the same audiences.

Second, they would have to deal with an ex-ternal business, since Brown outsources theday-to-day running of the shuttles. Third,they’d have to contend with a host of techni-cal challenges: processing (roughly) real-time data, handling absent data, working withugly, real-world data files (such as road data-bases), programming for future changes(such as new buildings, or new names for ex-isting ones), offering different interfaces fordifferent users (especially technologically na-ive ones), building simulators, and so on.Fourth, they’d have to make this all afford-able. And finally, they would get to play withcool devices.

Cool devices? My TAs, Nathan Weston andJohn Goodwin, and I (especially Nathan) dida fair bit of research over winter break to de-termine that we’re at an interesting techno-logical cusp. While there are many devicesthat would help tackle such a problem, wewanted something as flexible and extensibleas possible—i.e., something programmable.Cell-phone technology has, fortunately, justgotten to this point. The phones have GPS re-ceivers, in part to comply with new E-911laws. This is fine for telling the phone whereit is, but would be useless if we couldn’t noti-fy a base station. The phones also have wire-less Internet access—also handy, but we’dstill need to get data from the GPS to the In-ternet. The crucial cog in this machine is thatthe same phones also have Java Virtual Ma-chines, with the ability both to use the Inter-net connectivity and, more importantly, anAPI for accessing the GPS data.

We took a while getting the class to theproject. Since they were going to deal with areal customer, we had them conduct mock in-terviews with the course staff, masqueradingas employees of a library in a small town inTexas (an excuse for me to trot out my cow-boy hat and bolo tie). You can read the exer-cise setup from the course’s Web page—search for “Brown CS190 spring 2004”.(Wondering about the choice of town? Searchon “Archer City Texas Larry McMurtry”.)Some of the course staff played technologicalignoramuses and even Luddites, and it wasinteresting to watch some students reallystruggle with this. So this was useful prepara-tion for the main project.

I had a few desires for what should happenduring the project. I wanted to see the re-quirements, or at least their focus, change; I

conduit! 33

wanted administrative structures to get in theway; I wanted team personnel to alter alongthe way; and I wanted to stress the impor-tance of prototyping.

In case these didn’t occur naturally, thecourse staff had plans for injecting them(even if a little artificially) into the process.

Fortunately, we never needed to intervene onany of these accounts. The requirementschanged because interviews with the dis-patchers revealed that the real-time mapwasn’t their greatest need. In fact, we foundthat they had much more mundane con-cerns—tasks such as producing audit trailswere far more bothersome and time-consum-ing. We also found later on that drivers weresometimes frustrated with getting insufficientor incorrect information over their cracklytwo-way radios; fortunately the cell phoneshad usable displays, so the group eventuallyadded address transmission as a feature.

The administrative structures became a con-cern because of the separation betweenBrown and the contractor running the shut-tles. The contractor naturally wanted to pro-tect his drivers’ time, so students were forcedto route questions through the manager. Theylater worked around this by informally inter-viewing the drivers while riding with them inthe shuttles at night.

The most important personnel change I want-ed to make was to reduce the size of one ofthe teams. As luck would have it, the veryweek I wanted to scale down that team, a stu-dent in that team dropped the class. I initiallyannounced this to the class by saying we hadfired him. A few of them looked terrified un-til I explained I’d been joking.

Finally, the need for prototyping came natu-rally. Because the students had had only lim-ited access to the drivers, they knew that theirrequirements were quite incomplete. Theyquickly realized that it was crucial to deploya prototype both to improve their understand-ing of the requirements and to use the feed-back phase as a covert way of interviewingthe drivers some more.

The class divided into three teams. The phoneteam of five people was responsible for figur-ing out the intricacies of the devices and actu-ally making them work as advertised. (Iconfess to thinking this would be the weaklink in the chain—not because of the studentsbut because of the novelty of the technology.The students did have to put in an extraordi-nary effort to overcome buggy systems, poorand missing documentation, and CIS’s andNextel’s technical support and bureaucracy,but they made it work!) The server team ofseven was responsible for receiving messages

from the phone, generating logs, and present-ing vehicle status and real-time map informa-tion. Finally, a five-member AI teamcoordinated with the server to generate routeinformation and estimate rider wait times.Each team was organized into a hierarchy re-sembling a small company.The team had an overall administrative head,as well as a head of quality assurance. In ad-dition, specific students in each team weretagged with responsibility for documentationand team-specific tasks.In addition, the class project was headed bytwo managers: Peter Woo, the product man-ager, and Daniel Stowell, the head of integra-tion and product-level quality assurance.The students did produce a prototype, thougha bit later than we’d planned.Each deployment of the software was greetedwith excitement.I had planned to both visit the dispatching of-fice and ride in the vans, but decided to stayaway the first day. This was fortunate.Due to outdated initial data, the systemcrashed when it was first deployed.Peter Woo joked that some students agedhours while fixing it and, had I been there,they’d have aged years. (What they didn’tknow at the time is that I was monitoringprogress through the Web-interface.) But af-ter this was fixed, the system ran splendidly.Sure, there were some small bugs, and the in-terface needed several improvements—butthat’s the point of prototyping, right? In an-other month, they had these bugs worked outand the result is a very professional product.I’ve worked students pretty hard in coursesbefore, but this group surpassed all demands.One night midway through the semester, theyhad a meeting that grew into a three-hour jamsession. They weren’t arguing about person-alities; they were arguing about the best wayto build the product.That kind of enthusiasm is, of course, aBrown characteristic, but this group of stu-dents seemed especially motivated. In end-of-semester surveys, they admitted that hav-ing a concrete client expecting a product fromthem, and knowing that it would be used byreal users (especially their fellow students),was an exceptionally strong motivation. Atthe end of the semester, Peter produced a CDwith photographs culled from class sessions,group meetings and parties from across thesemester. It was a touching gesture.I owe thanks to several people who gave free-ly of their time. Steve Reiss and David Laid-law offered a lot of advice from having taught

conduit! 34

CS190 in the past. Ugur and John Jannottihelped me better understand wireless tech-nologies. Amy Greenwald taught the entireAI team, and their work and enthusiasm wasa testament to her training. Abbi Rider got usstarted on this project and supported its de-velopment; the folks at Nextel gave us free-bies to keep us well under budget. Thanksalso to the department’s technical staff, forsetting up a server and supporting it; to BobPerreira and Robin Carillo in Telecommuni-cations at Brown, who helped us get cellphones and resolved administrative issueswith Nextel; and to Alan Usas of CIS.There’s a happy sequel to this course. Agroup of students has spun off a company,East Transit Technologies, Inc., to market thissoftware to other universities and, they hope,

later expand into other markets that can uti-lize such tracking. They tell me they’ve justset up their first demonstration for a potentialclient outside Brown, so there’s great excite-ment about their prospects.

East Transit Technologies is named for thestreet on which they’re living in Providencethis summer. But perhaps there’s a more cos-mic connection.

Transit Street in Providence gets its name notfrom some public transportation connection,but rather from popular enthusiasm for obser-vation at the then-fledgling Rhode IslandCollege of the 1764 Transit of Venus. Perhapsit’s fitting that the company formed on therare repetition of the same astral phenome-non, 240 years later.

Probably you are tired of rubber chicken sto-ries, but they are like kittens, they seem tolead to yet more rubber chicken stories, and I

could not resist this one.

Eugene,

I loved the recent Con-duit stories about therubber chicken tradi-tion. I had actuallybeen thinking aboutthat recently, wonder-ing exactly how thetradition started—nowI know. My own rub-ber chicken is careful-ly stored away in thecloset, still in fineshape after almost 14years.

Last summer, my daughter—then 3.5years old—saw it and asked if shecould play with it. I explained to herthat it was very special and not a toy.She asked when she could get her ownrubber chicken. I said she had to finishpreschool, then elementary school,then middle school, then college, thengraduate school, then get her PhD andthen she could have a rubber chicken.

Nothing more was mentioned forseveral months. Then in November,our family was sitting around the din-ner table discussing jobs and retire-

ment. My daughter asked if she couldretire. I said no, she had to have a jobbefore she could retire. She askedwhen she could get a job. I said firstshe had to finish preschool, then ...When I got to the part about the PhD,she got all excited and said, “then Ican get my rubber chicken!”

Who knows—maybe there’s a wholenew generation of rubber chickens onits way.

Randy Calistri-Yeh Ic.U. (IckenchayUbberray) 1990

(give me a break— I never studiedLatin...)

Randy, when your daughter applies to ourPhD program, make sure she puts on the ap-plication that she is a legacy rubber chicken.

Off and on during my life I have created “art”of various forms. When I first arrived atBrown I produced an abstract “painting”made of various common materials all heldtogether with glue, and it has been in my of-fice ever since. In my current CIT office it isright behind my desk, so anyone talking tome has to be staring at it, but to my amaze-ment large numbers of people never notice it.Those who do have a variety of opinions,mostly negative. The mostly negative groupstays mostly silent, of course, with the excep-tion of Andy van Dam, who tells me what hethinks of it about every other time he comes

CHARNIAK UNPLUGGED

Eugene delivering his “Snapshots of the Department’s Past” talk at the 25th anniversary

banquet

conduit! 35

to my office. I was going to get a quote fromAndy for this article, but I decided not to. Onthe topic of my painting his creativity seemsto disappear, and besides, the commentwould be unprintable. Andy’s evaluation,however, was nothing compared to anothervisitor to my office. After our new caretakerhad been in my office several times, he askedme, in all seriousness, if anyone had calledbuildings and grounds about the problemwith the wall behind my desk. I did not havea smart reply.

Many a conduit! ago I mentioned that when Iwas spending four days a week away frommy family on a sabbatical at Johns Hopkins,my wife gave me a 1000-piece puzzle of a po-lar bear on snow. After reading this, SuziHowe loaned me one of her jigsaws, this oneof hundreds of marbles spilled all over. Thispuzzle sat in my office unsolved for a fewyears until I decided I was very unlikely to getaround to it and I should give it back to Suzi.

Instead, however, I put it in the grad studentlounge. Lo and behold, it was all put togetherthe very next day. I then returned it to Suzi,along with the story. She noted the similarityto the scene at the start of ‘Good Will Hunt-ing’ when Will solves the math problem atMIT. (But I don’t think the movie rights aregoing to be worth much.)

I was talking to Trina Avery and she told methat she finally broke down and started put-ting her pills for the week in a seven-com-partment pill box so that she could trackhaving taken them. She (quite rightly I think)saw this as a geriatric thing to be doing, butwhen she mentioned this to Tom Doeppner hesaid that this was simply “pre-fetching” andthus perfectly OK. This led her and Tom to adiscussion of other CS terms that applied todaily life, but unfortunately I cannot remem-ber what they were. (I went to ask Trinaabout this, only to discover that we were hav-ing simultaneous senior moments.)

At any rate, Trina asked me if I wanted to usethis for my conduit! column. I said “no”, butI had a hard time thinking of why not. Even-tually I came to an important realization. My

Department of Computer ScienceBrown University

Box 1910, Providence, RI 02912, USA

conduit!

conduit! 36

NON-PROFITU.S. Postage

PAIDProvidence, 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 Science

Box 1910, Brown UniversityProvidence, RI 02912

Fax:.401-863-7657Phone:.401-863-7610

email:.sjh@cs.brown.eduWWW:.http://www.cs.brown.edu/

publications/conduit/

conduit! article was not really about the de-partment, it was about ME, and secondarilyabout ME and the department. Since Trina’sstory was not about me, it did not qualify.For example, Suzi told me a very nice storyabout Shriram for my column. It seems thathe decided to teach on the CS balcony oneday, but was unable to remove the “port-able” whiteboard from Lubrano. Instead, hefound that if he lowered the window blinds,the windows onto the balcony made a per-fectly usable blackboard—see the picturesabove. I was planning on rejecting this sto-ry, too, for the same reason. Having had this

realization, I went back to Trina and ex-plained why I had said “no”. Trina said toobad, it’s a nice story, but I responded that nowI could use it, because now it was no longer astory about Trina and Tom discussing the useof CS terms, but a story about me having a re-alization about a story about Trina and Tom...Trina responded, “Ah, recursion!”

When this conduit! was in proof, ShriramKrishnamurthi pointed out that it’s not recur-sion, it’s self-reference. But this is my col-umn, so the word means what I want it tomean; as Humpty Dumpty says, ‘the questionis who’s to be master, that’s all.’

the conduit! team salutes the department’s 25th anniversary. l to r: Jeff Coady, Suzi Howe,

Eugene Charniak, Trina Avery