the The Magazinelink. of the Carnegie Mellon University School of Computer Science spring 2013 issue 7.1

Smashing the stereotypes

We look at ways to make women and other under-represented groups

feel more welcome in the world of computer science

also inside: Class of ’79 alumna shares 2012 Turing Award Does your typing rhythm hold clues to your health? Research Notebook: Low-literacy communities connect via social media thelink. The Link provides a mosaic of the School of Computer Science: presenting issues, analyzing problems, offering occasional answers, giving exposure to faculty, students, researchers, staff and interdisciplinary partners. The Link strives to encourage better understanding of, and involvement in, the computer science community.

Editor-in-Chief Randal E. Bryant

Editor Jason Togyer

Contributing Writers Tina Carr, Ken Chiacchia, Mark Dorgan, Meghan Holohan, Mary Lynn Mack, Linda K. Schmitmeyer, Byron Spice

Photography Ken Andreyo and Tim Kaulen, Carnegie Mellon University (pp. 1, 2, 16 and 37), John W. Barna (p. 14), Jason Dorfman, CSAIL/MIT (p. 33), National Robotics Engineering Center (p. 34, bottom), SURTRAC (both, p. 5), Tartan Rescue Team (p. 34, top) and University of Pennsylvania Archives (p. 38). Other images courtesy John Anderson (p. 35, top), Justin Cranshaw (p. 35, bottom right), Alon Lavie (p. 8), Gary L. Miller (p. 36), Agha Ali Raza (p. 25), Julian Shun (p. 35, bottom left), Pradeep Sindhu (p. 30), Séan Slattery (p. 32), Stefanie Tomko (p. 31) and Daniel H. Wilson (p. 13).

Graphic Design Communications Design & Photography Group

Office of the Dean Gates Center 5113 Carnegie Mellon University 5000 Forbes Avenue Pittsburgh, PA 15213 Randal E. Bryant, dean Tina Carr (HNZ’02), director of alumni relations Philip L. Lehman (CS’78,’84), associate dean for strategic initiatives Byron Spice, director of media relations Jason Togyer (DC’96), editor

Phone: 412-268-8721 Email: [email protected] Web: link.cs.cmu.edu Facebook: facebook.com/SCSatCMU Twitter: twitter.com/SCSatCMU

Carnegie Mellon University does not discriminate in admission, employment, or administration of its programs or activities on the basis of race, color, national origin, sex, handicap or disability, age, sexual orientation, gender identity, religion, creed, ancestry, belief, veteran status, or genetic information. Furthermore, Carnegie Mellon University does not discriminate and is required not to discriminate in violation of federal, state, or local laws or executive orders.

Inquiries concerning the application of and compliance with this statement should be directed to the vice president for campus affairs, Carnegie Mellon University, 5000 Forbes Avenue, Pittsburgh, PA 15213, telephone 412-268-2056.

Obtain general information about Carnegie Mellon University by calling 412-268-2000.

Produced for the School of Computer Science by the Communications Design & Photography Group, April, 2013, 13-328.

©2013 Carnegie Mellon University, all rights reserved. No part of this publication may be reproduced in any form without written permission from the Office of the Dean, School of Computer Science.

2 contents 2 dean’s message

3 on campus

14 in the loop

16 cover story

25 research notebook

29 alumni

35 scs news in brief

39 screenshot

40 then and now

41 calendar of events

on the cover Cover illustration by Jordan Bush; photography by Tim Kaulen the the . link

1 Dean’s message

Reflecting on the Cohon Era Jared Cohon will be stepping down as president of Carnegie Mellon on June 30 after 16 years of service. It’s remarkable how the School of Computer Science has flourished during that time period. Our total student enrollment, including undergraduate, masters and Ph.D.’s, has almost doubled to its current level of 1,600. Our research budget has nearly tripled, to its current level of $110 million. We have greatly expanded our physical footprint as well with the opening of Newell-Simon Hall, the Gates and Hillman Centers, and several renovated office buildings near the Pittsburgh campus on Craig Street. President Cohon has been an enthusiastic supporter of computer science throughout his tenure. By all indications, Subra Suresh will be a strong supporter of computer science. Suresh comes to us with impressive credentials, including his current position as director of the Dean Randal E. Bryant National Science Foundation, and before that dean of the College of Engineering at MIT. We look forward to welcoming him as the ninth president of Carnegie Mellon University on July 1.

Randal E. Bryant Dean and University Professor School of Computer Science

The university’s new president received an Science Foundation. Ray Lane, chairman of CMU’s enthusiastic welcome on Feb. 21. Among those board of trustees (far right), calls Suresh “uniquely greeting Subra Suresh and his family in the qualified” to lead the university in part because University Center’s Rangos Hall was the Home of his commitment to technology research Exploring Robot Butler, or “HERB,” a Robotics and education. Institute creation designed to help people with Also shown are Suresh’s wife, Mary Delmar, physical challenges live independently at home. the former director of public health for the town Suresh is former dean of engineering at MIT of Wellesley, Mass., and their daughters Meera and currently serves as director of the National and Nina. 2 the link. 3

Born in Rüsselsheim, Germany, studied Pfenning and mathematics science computer the Technical at University With Darmstadt. the support a Fulbright of he Scholarship, Carnegie attended where Mellon, he earned a degree master’s Pfenning Byron Spice is director of media relations for the School School the for relations media of is director Spice Byron rank rank of Computer Science. Computer of — F of Computer Sciencein 2002. Computer of international of boards served has numerous He on academic and journals research organizations, professional conferences. in mathematics in 1981 and a Ph.D. in mathematics in mathematics in a Ph.D. 1981 and in in mathematics scientist, a research as CSD joined subsequently 1987. He in 1999 professor associate as appointment an receiving of served director in 2002. He CSD as fulland professor dean associate as 2008 and 2004 to from programs graduate Science Computer the School of for education graduate of appointment adjunct an alsoholds 2010. He 2009 to from Philosophy. of in the Department the Max- at scientist been has a visiting Pfenning Science in Saarbrücken, Computer for Planck-Institute the Technical at Fellow Alexander-von-Humboldt an École at professor a visiting and Darmstadt, University the Herbert won He INRIA-Futurs. and Polytechnique in the School Excellence Teaching for Award A. Simon

Spice campus yron Byron on formulating and solving problems. solving and formulating computer scientists have developed unique ways of of ways unique developed have scientists computer concepts of computational thinking—the idea that computational of concepts emphasis on parallel computation and incorporate incorporate and parallel computation on emphasis to developing reliable software. They place greater greater They place software. reliable developing to the past several years, reflect a more rigorous approach approach rigorous more a reflect severalthe past years, other new courses, which have been implemented over over been implemented which have new courses, other introductory computer science curriculum.and This computer introductory Imperative Computation, that is part of CSD’s updated updated CSD’s part is of that Computation, Imperative mathematical logic in computer science. This includes includes This science. logic in computer mathematical of Principles a new developed course, Pfenning Pfenning’s research focuses on applications of of applications focuses on research Pfenning’s security. computer ensuring computer science master’s program.” science master’s computer logics for and programs computer about reasoning on the committee that designed our recently introduced introduced recently designed our that the committee on for systems languages, programming the design of progress of around 150 Ph.D. students. He also served He students. 150 Ph.D. around of progress computer science Ph.D. program, keeping track of the of track keeping program, science Ph.D. computer said. “At the graduate level, he served the he of level, director as the graduate “At said. as computer systems and computer graphics,” Bryant Bryant graphics,” computer and systems computer as such as mathematical logic, to the very applied, such such the logic, very to applied, mathematical as such members. Established in 1965, CSD was one of the of one was in 1965, CSD Established members. thevery from theoretical, courses, of array amazing will lead a distinguished faculty of more than 70 than more will faculty of lead a distinguished an taught has Frank level, the undergraduate “At As head of the Computer Science Department, Pfenning Science Pfenning Department, the Computer head of As departments. science computer first world’s shown great care for the welfare of students. of thewelfare for care great shown involved in the school’s educational activities and has has activities and educational in the school’s involved Dean Randy Bryant said Pfenning has been has heavily Pfenning said Dean Bryant Randy become head of Microsoft Research International. SCS International. Research Microsoft head of become He succeeds Jeannette Wing, who stepped down to to down who stepped Wing, succeedsJeannette He Science Department. science curriculum, is the new head of the Computer science curriculum, the Computer the new is head of Carnegie Mellon University’s introductory computer computer introductory University’s Carnegie Mellon science who played a key role in the recent update of of update in the recent role a key science who played Frank Pfenning (S ’81, ’87), a professor of computer computer of (S ’81, ’87), a professor Pfenning Frank By

New CSD head had role role head had CSD New in program’s design in program’s on campus All in good time Traffic-sensitive signals developed at the Robotics Institute are saving fuel, time and drivers’ nerves

By Linda K. Schmitmeyer According to Smith, similar software is already being used to manage traffic flow onto along arterial roadways— Gregory Barlow knows a lot about traffic, and not just how surface roads with a strong traffic flow in one direction long it takes to commute from his home in Squirrel Hill and some side streets. With these systems, though, to his office in Newell-Simon Hall, where he works with information is typically collected and sent to a central CMU research professor Stephen Smith on tackling traffic location, where data are analyzed and adjustments to the congestion in urban areas. signaling are determined. Barlow (CS’11), a post-doctoral researcher in “It can take anywhere from five to 15 minutes the Robotics Institute, knows traffic problems (to reprogram the signals), depending on are nothing new: The early Romans wrestled the system being used,” Smith says. with them two millennia ago by banning “The beauty of our approach is that it is wagons on their roads during certain times decentralized, which makes it inherently of the day. He also knows they’re costly: In scalable, in principle.” 2009, drivers in 439 urban areas in the United States traveled 4.8 billion hours longer and There also are decentralized systems used on purchased 3.9 billion more gallons of fuel arterial roads, says Smith, but they operate on because of traffic snarls, for a total congestion the assumption that there is a dominant flow cost of $115 billion. of traffic that does not change. And all this comes from Barlow, a man who admits, SURTRAC, in contrast, is designed to discover the “I don’t really like to drive.” dominant flow of vehicles through an intersection and automatically adjust the signaling. It is designed to operate Maybe that’s why he’s working to improve the driving within urban grids, where the volume and direction of experience in Pittsburgh, along with CMU colleagues traffic can change throughout the day. Xiao-Feng Xie, a research associate, and Zachary B. Rubinstein, a senior systems scientist. They are part of a At the East Liberty intersections, cameras take continuous team led by Smith, director of the Intelligent Coordination shots of the traffic, which allows SURTRAC to create a and Logistics Laboratory in the Robotics Institute, where schedule for moving vehicles through the intersections they developed a “smart” traffic signal that works in real- in the most efficient way possible. Each intersection also time. Smith has dubbed the project Scalable Urban Traffic communicates via a fiber optic cable (or, in the case of Control (SURTRAC). one East Liberty intersection, a wireless signal) to its downstream neighbors as to what the projected outflow Today, most signals operate by pre-setting the timing on from its signal will be. The neighboring signals do the same the green light for several different periods throughout thing, and together they create a communications network the day, depending on the number of vehicles expected that’s akin to having the watchful eyes of traffic police at to travel through an intersection. As part of the Traffic21 every intersection. Initiative in Carnegie Mellon’s H. John Heinz III College, the researchers developed software that allows a signal More impressive, though, are the net results. At the East to respond to traffic as it is happening. Last spring, they Liberty intersections, the research team found that the installed adaptive signaling systems at nine intersections wait time for people driving through the grid was reduced in Pittsburgh’s East Liberty neighborhood. by 40 percent. Travel time was reduced by 26 percent and projected vehicle emission by 21 percent. The researchers “Our system watches actual traffic flow through the obtained the results by completing “before” and “after” cameras at each intersection and dynamically adjusts drives along 12 predetermined routes through the nine the green timing periods on a second-by-second basis,” intersections. They did the drives during four set times Smith says. throughout the day; GPS tracers on their cell phones

4 the link. 5

—Linda K. Schmitmeyer is a writer and editor in Valencia, Pa. Pa. Valencia, in editor is and a writer K. Schmitmeyer —Linda byline. Link first her is This which has already received approval to install SURTRAC- to approval received already which has East LibertyPenn from along signals eastward supported will be intersections more Nine Avenue. Fifth to Avenue also is signaling 2013. Adaptive theadded grid to during to avenues Center and Baum along being considered the pilot. for Craig Street in Oakland, creating what Smith likes to call to likes Smith what in Oakland, creating Street Craig should town into way a faster but Rome,” to leading drivers. please many the from with 2009 in funding launched was Traffic21 from Traffic21 to Grants Foundation. Henry L. Hillman the and from Project Breathe Heinz Endowments’ The the funding provided Foundation Mellon King Richard “a virtual corridor to Downtown.” Downtown.” to virtualcorridor “a may Pittsburgh into corridor A computer-enhanced roads “all of accomplishment the Roman match not

monitor traffic in real time. “What’s exciting is that this isthat exciting “What’s real in time. traffic monitor place to place,” Cunningham says. “I can see how (the see can how “I says. Cunningham place,” to place munici- that is Cunningham, says advantage, Another be can installed to system a SURTRAC equipment, tion forthe engineer traffic municipal Purcell, Amanda the project, expand to way under talks says are She competitive than being able to move people easily from easily from people move to being able than competitive in real estate on impact an have flow) easewill in traffic East Liberty.” the technologyincrementally. implement palities can - detec been include has to upgraded intersection an Once little off bite can Cities so sois cheap. powerful and available.” become funds as a time, at chunks in East project thepilot that says Pittsburgh, of City a “establish willand help Liberty fine” “is working the signals operate.” how baseline for “There is no more important thing to make a community a community make to thing important more “Thereno is Liberty intersections as a boost to the local economy. a boost as Liberty the local to economy. intersections Development Inc., sees the ease in congestion at the East at sees the ease Inc., in congestion Development Nate Cunningham, director of real estate for East Liberty for estate real of director Cunningham, Nate implications,” Biehler says. says. Biehler implications,” Transportation. “This (technology) will have far-reaching “This far-reaching (technology) willhave Transportation. to discuss SURTRAC with his state’s Department of of Department state’s his with discuss SURTRAC to Secretary Ananth Prasad invited Biehler and Smith Smith and Biehler invited Secretary Prasad Ananth Highway Transportation Officials, Florida Transportation Transportation Officials, Florida Transportation Highway the annual meeting of the American Association of State State of Association the American of meeting the annual States will be interested in SURTRAC. In November, during during November, In in SURTRAC. willbe interested States more data, he expects other cities throughout the United the United expects he throughout cities data, other more be done” by expanding the pilot grid in an effort to obtain obtain to grid effort in an the pilot expanding by be done” Although Biehler says that more “robust testing needs to testing “robust more that says Biehler Although Transportation. Transportation. and former secretary of the Pennsylvania Department of of Department secretary the Pennsylvania former of and director of CMU’s University Transportation Center Center Transportation University CMU’s of director were as good as they were,” says Allen Biehler, executive executive Allen Biehler, says good as they as were,” were “When I saw the results, I was really shocked that they that shocked really I was the results, I saw “When federal transportation and economic development funds. development economic and federal transportation a regional planning agency that administers state and and state administers that agency planning a regional frequently by the Southwestern Pennsylvania Commission, Commission, Pennsylvania the Southwestern by frequently Smith says the methodology says basedSmith is a modelused on the differences. the differences. allowed them to gather the data necessary the data calculate to gather themto allowed on campus Decoding a cyber-fingerprint

Your typing rhythm holds clues to your identity, and maybe even your future health

By Mary Lynn Mack keys are pressed, the lab also collects information about The blink of an eye takes 300 to 400 milliseconds. It takes user demographics, hand geometry, hand and body less time than that—about 90 milliseconds, on average— posture, and movement. to press a computer key while typing. Could your password be customized? Individual keystrokes as well as the “rhythm” of typing As a result, his team is close to determining what password a word, sentence or document are forming the basis of a they could assign to someone that could be distinguished cyber equivalent to handwriting or fingerprint analysis. not just because of the mix of numbers, letters and In a small laboratory on the eighth floor of CMU’s Gates punctuation marks, but because of the way the user types Center for Computer Science, Dr. Roy Maxion and his the password—their individual typing rhythm. In other team have developed the ability to capture and analyze words, “can we customize your password so that your those 90-millisecond keystrokes and read the clues typing style would be best conveyed by that particular encoded in them. Called “keystroke dynamics,” Maxion password?” Maxion says. For instance, for someone with believes the research may have the potential to not only short, thin fingers, a password with characters that span the change the world of cybersecurity, but also to identify keyboard would be more difficult to type than one that had individuals with musculoskeletal diseases, neurological characters grouped together. Conversely, for an individual disorders, cognitive decline and acute stress. with thick fingers, a combination of characters clustered in the center of the keyboard would require more effort and Maxion, a research professor in the School of Computer deliberate attention. Science, became interested in keystroke dynamics about five years ago when he was asked to review a journal article There are “lots of thorny little problems” that can affect a on the subject. Although research into keystroke dynamics person’s typing pattern, such as stress, fatigue, illness, injury began 30 years ago, some of the most significant progress or substance use, Maxion says. But just as a person can be has been made in the last decade or so. The topic was new identified by his or her gait, even when they’re carrying a to Maxion, and it interested him. Within a short time, heavy object or wearing different shoes, Maxion says that Maxion was working on a National Science Foundation- someday, software may be able to identify someone’s core funded project to determine if someone could be identified typing rhythm despite a slight deviation from the norm. Finding that core typing rhythm, he says, may unlock the from his or her typing style. doors to reliable, real-world applications. “My position is that no one has demonstrated that it isn’t Although cybersecurity continues to be a major focus, possible,” he says. Maxion says the research has implications in health care. Cybersecurity is Maxion’s speciality. A member of the Would it be possible, for instance, to detect whether National Academies’ Committee on Future Research Goals someone’s typing rhythm is changing because of illness and Directions for Foundational Science in Cybersecurity, or injury? Maxion’s research suggests that changes in Maxion serves as vice chair of Professionals for Cyber typing rhythm could provide early warning signs when Defense and director of CMU’s Dependable Systems someone is beginning to suffer from musculoskeletal Laboratory, and has also studied intrusion detection. His problems such as carpal tunnel syndrome, digital flexor early interest in keystroke dynamics focused on ways tendonitis and arthritis. of lowering the error rate when analyzing keystrokes in Maxion recently partnered with Nancy Baker, an associate applications such as Web-based financial transactions— professor of occupational therapy at the University Maxion’s group was able to improve the accuracy to 99.97 of Pittsburgh’s School of Health and Rehabilitation percent in 10 keystrokes. Sciences. Baker has been conducting research on detecting Maxion’s lab is now able to measure typing rhythms on musculoskeletal disorders through keyboard use since 2008. a scale far more precise than any other known system “In my experience, as people develop illness of or injuries to available. In addition to capturing the length of time the the soft tissue, the way they perform certain tasks changes, 6 the link. 7

—Mary Lynn Mack is a Washington, Pa.-based freelance freelance Pa.-based is a Washington, Mack Lynn —Mary and interesting.” and be used as part of the diagnostic process for these diseases. for be used the process diagnostic part as of and detection early disorders, musculoskeletal with As benefit to chance a better patients give could diagnosis baseline information obtain we “If treatment. early from specific diseasesor without and with people enough on the separate to be to able going are then we disorders, to correspond that rhythms the keystroke of characteristics says. Maxion those diseases disorders,” and those what know to too early while it’s that notes He those lead be, learning could signs might characteristics by the determine, can that software of the development to a to a susceptibility types, individual if they an have way particular disease disorder. or collection data careful Precise, keystroke of the field advanced CMU has that says Maxion team His methods. research specific its because of dynamics the time measure can that mechanism timing own its built available— previously speeds not at keys press takes to it the they And standardized levels. sub-millisecond experimental tight incorporate to their data collection of asked were 51 subjects During the initialstudies, control. the same using desk, the same at room, in the same sit to type 10-character the and same computer, keyboard and sessions. eight 400 times over password most the is possibly setting standardized controlled, The beingdone the research distinguishes that thing important dynamics, keystroke of studies other Many team. Maxion’s by conditions real-world under data gathered have says, he type, variables—keyboard themany for controlling without not may or may fluency—that language posture, typing to doing that likens Maxion rhythm. typing affect someone’s the says chemistry and dirty tubes, with experiments test has team Maxion’s the map.” over “all are datasets resulting researchers other for online dataset standardized its published work. usetheir in own and download to it looks dynamics, keystroke approach first you “When and “The some goeskey says. he down, simple,” pretty start to when you But up. goes back it millisecondslater, need be to that subtleties of a lot are there it, into get hard— it makes what That’s with. dealt recognized and writer. This is her second Link byline. Link her second is This writer. esting typing to predict illness predict to typing esting rhythm, which suggests keystroke dynamics could one day day one could dynamics keystroke which suggests rhythm, dementia can be detected through changes in typing in typing be changes can detected through dementia indications that early signs of Alzheimer’s disease and Alzheimer’s of signs early that indications help identify neurological disorders. There also are some some are also There disorders. identify neurological help control, and which is part of a battery of tests used to tests of a battery part which is of and control, Tapping Test, which assesses motor speed and motor speed motor and which assesses motor Test, Tapping type of results generated by the Halstead-Reitan Finger Finger the Halstead-Reitan by generated type results of Maxion sees and the between a similarity those findings Maxion recently been diagnosed with right temporal lobe epilepsy. been diagnosed temporal right with recently more investigation, the team learned that the subject had the subject learned that theteam investigation, more milliseconds, about half the average of other subjects. With With subjects. other of half the average milliseconds, about Additionally, the subject held his keystrokes for just 46 just for keystrokes his held the subject Additionally, a very unusual, distinctive typing pattern, Maxion says. says. Maxion pattern, typing distinctive a very unusual, recent test, the researchers identified a subject had a who identified the researchers test, recent keystroke dynamics and neurological issues. During one During one issues. neurological and dynamics keystroke The CMU team also is looking at connections between at connections team also looking is CMU The difference, it would be a substantial savings,” Maxion says. Maxion savings,” substantial would a be it difference, “If by using our technology we could make even a 1 percent even a 1 percent make could technology we our using by “If productivity, at between $45 billion and $54 billion per year. per year. $54 billion and between $45 billion at productivity, U.S. economy of such disorders, including lost wages and and wages lost including disorders, such of economy U.S. and the Institute of Medicine estimated the cost to the to the cost estimated Medicine of the Institute and months. A 2001 report from the National Research Council Research the National from A 2001 report months. tunnel syndrome at some point during the previous 12 the previous during point some at syndrome tunnel of U.S. adults aged 18 to 64 reported suffering from carpal from suffering 64 reported 18 to aged adults U.S. of Disease Control and Prevention, an estimated 3.1 percent 3.1 percent estimated an Prevention, and Disease Control common and costly. In 2010, according to the Centers for for the Centers to according 2010, In costly. and common reverse. Work-related musculoskeletal disorders are both are disorders musculoskeletal Work-related reverse. a disorder, Baker says, the more difficult it is to treat or treat to is it difficult themore Baker says, a disorder, and prevent further damage. The longer a patient has a patient longer The further damage. prevent and Early detection of such disorders can improve treatment treatment improve can disorders such of detection Early at risk of developing a musculoskeletal disorder. disorder. a musculoskeletal developing of risk at determine which postures, angles and positions put a user put positions and angles which postures, determine at specific moments (kinematics), the two hope to be able to to able hope be two the (kinematics), specificmoments at measurements to determine the location of joints in space joints of the location determine to measurements studies, which include visual observations as well as well visual as observations which include studies, By pairing Maxion’s precise measurements with Baker’s Baker’s with measurements precise Maxion’s pairing By T each hand to type. to each hand and fingers in a fixed position, and use only one finger on finger one only and use in a fixed position, fingers and rest, move their hands to strike keys, maintain their wrists strike maintain keys, to their hands move rest, more likely to use high-force keystrokes, not use a wrist not keystrokes, use high-force to likely more says. For example, subjects with rheumatoid arthritis are arthritis are rheumatoid with subjects example, For says. as they adapt and respond to changes in the tissues,” Baker in the tissues,” changes to respond and theyas adapt Decoding a cyber-fingerprint on campus Sweating the details

SCS startup Safaba develops smarter translations for specialized clients

By Meghan Holohan of a computer-generated translation phrase such as “from now officially Cologne,” but it doesn’t sound right. Say a marketing executive at a large company needs to translate new product information on her company The company exec who used an online, mass-market website from English to French within a few hours. She’s translation program to translate her website from English not fluent enough in French to do it herself, so she has to French could end up with text using either slang several options—she can hire a human translator or a (informal language instead of formal) or terms from the translation agency. She can also use a readily available wrong domain or industry. Lavie uses the example of the online translation program, such as those offered by English word “tablet.” Imagine translating “tablet” on the Google and Microsoft’s Bing. website of a computer company as if it were a medicine “tablet” or pill, he says. Translating an entire website using human translators would be time-consuming and costly, and the commonly Unlike free services that aim for very broad audiences available free online translation services have drawbacks. but provide somewhat crude translations, Safaba targets large organizations—such as companies in the Fortune “When people think of 500—that need fast but high-quality translations of high language technology and volumes of documents, often incorporating very specific translation, most people corporate language. (think) of free services, like Google, that are basically “We use machine-learning to give them what they need,” designed to support a very Lavie says. broad range of individual The global market for translation services is estimated at users who might want $30 billion annually, with machine translation currently to translate anything,” Alon Lavie accounting for a fraction of this market. As demand for says Alon Lavie, research large volumes of content translation increases and the professor at CMU’s Language Technologies Institute technology improves, experts believe that demand for and president and CEO of the startup Safaba Translation machine translation will continue to expand. Solutions, which provides specialized computer-based machine translation software for global corporations. “The industry itself is growing,” says Olga Beregovaya, a vice president at Welocalize Inc., which is using Safaba’s “What Google and Microsoft are trying to do is build translation engine to provide specialized, on-demand systems that generate the best possible translation without translation services to major clients with global audiences, knowing anything specific about your company and the such as Dell and PayPal. specific terminology and other language characteristics that you and your company typically use,” Lavie says. While working on research projects as a faculty member While those translations are often understandable, he at the Language Technologies Institute, Lavie became says they’re not quite good enough for representing your involved in the Association for Machine Translation in the company in another market. Americas, or AMTA. Through AMTA, Lavie saw that the machine translation market needed programs that better To cover their bases as broadly as possible, Lavie says, free captured certain specialized nuances of language. translation services build their translations from a wide range of different resources. The resulting quality can vary Lavie teamed up with his former Ph.D. classmate Bob from one sentence to the next. Anyone who has translated Olszewski (CS’01), who was working at the time at a a webpage from another language to English, using a free different LTI spin-off company that creates software for online service, is familiar with awkward translations. Sure, language tutoring. The two began collaborating on a a native English speaker may be able to figure out the gist system that businesses could use to translate content into multiple languages.

8 the link. 9

access to machine-learning machine-learning to access at developed models in all possible contexts,” in all contexts,” possible Lavie says. having says Beregovaya translation that is correct correct is that translation To prevent errors from from errors prevent To always edit Safaba output,” output,” Safaba edit always is “The here says. she benefit than faster is post-editing that going public, human translators translators human public, going Beregovaya of Welocalize. “We “We Welocalize. of Beregovaya human translation, and cheaper. cheaper. and translation, human human.” as read doesn’t before content goes live, says says goes live, content before (it) good, is but output Safaba still have to get final approval approval final get to still have Carnegie Mellon gives Safaba an an Safaba gives CarnegieMellon edge over competitors. “Initially, we went with went we “Initially, competitors. over edge them because of CMU’s name recognition,” she says. says. she recognition,” name CMU’s them because of —Meghan Holohan is a Pittsburgh-based freelance writer writer freelance is a Pittsburgh-based Holohan —Meghan Link. The to contributor a frequent and Welocalize then held a “technology bake-off” a “technology then held betweenWelocalize services, translation and competing two and Safaba victor. the clear was Safaba where a project on bidding currently is Welocalize between translations provide would engine Safaba’s However, the reliance on a company’s own translations translations own a company’s on the reliance However, the translated If own. its of lead problems can to will models the program’s errors, include materials single a isn’t often “There too. those errors, “learn” languages. 35 different better, works it “If says. Beregovaya pragmatic,” are “We money.” also we save and money save clients our

phrases that confuse native speakers. native confuse that phrases translations it generates include fewer of those awkward those awkward of fewer include generates it translations learning from professionally translated documents, the documents, translated professionally from learning accurately, Lavie says. And because Safaba’s software is software because Safaba’s And says. Lavie accurately, the specific nuances of a particular company quickly and quickly of a particular company nuances the specific backend authoring technologies, allowing businesses businesses allowing technologies, authoring backend to language its adapt to how on programs Safaba’s train Safaba’s software can handle multiple languages and languages multiple handle can software Safaba’s other and systems management content into directly needed is to material of smallamount a relatively Only company’s unique language. unique company’s services its incorporate eventually to hopes the company are the original as texts translations live generate to being written. generated translations to respect the individual respect the individual to translations generated translations of new documents and adapt the adapt and new documents of translations highly reliable statistical models used to generate used models generate to statistical highly reliable the materials, sentence-by-sentence, creating creating sentence-by-sentence, the materials, corporate slogans. Safaba’s software then analyzes then software Safaba’s slogans. corporate such as technical descriptions or trademarked trademarked or technical descriptions as such approved translations of specialized of language translations approved Those documents already contain contain Thosealready documents in, for example, English and French. French. and English example, in, for humans—the same documents documents same humans—the already had translated by by translated had already materials that they’ve they’ve that materials provide Safaba with source source with Safaba provide programs, companies companies programs, To create custom translation translation custom create To Entrepreneurship.) Center for Innovation and and Innovation for Center Carnegie Mellon University University Carnegie Mellon School of Business to create the create to Business School of Center for Entrepreneurship in CMU’s Tepper Tepper in CMU’s Entrepreneurship for Center Olympus recently merged with the similar Don Jones Jones the Don similar with merged recently Olympus Since then, it’s grown to include 11 employees. (Project (Project 11 employees. include to grown then, it’s Since launched in summer 2009 with just Olszewski and Lavie. Lavie. Olszewski and just 2009 with in summer launched Technology Transfer and Enterprise Creation, Safaba was Safaba Creation, Enterprise and Transfer Technology program Project Olympus and CMU’s Center for for Center CMU’s and Olympus Project program With support from the SCS-based entrepreneurship the SCS-based from entrepreneurship support With on campus Speaking of security Voiceprint authentication a la “Star Trek” could be poised to become reality

By Ken Chiacchia

Now, what was that password? This was the system that needed eight digits, right? Did it require symbols, or just letters and numbers? Today’s password-driven electronic security systems have a face only a system security administrator could love. Better to use some kind of biometric—something physically a part of us, marking us as ourselves. Some systems already use fingerprints. What about a voiceprint? Wouldn’t it be great if our computers could recognize us, the way the Enterprise recognized Captain Kirk saying, “Zero-zero-zero destruct zero?” Simply recognizing human speech isn’t all that hard for today’s computers. Take Apple’s voice-activated Siri assistant, for example. Matching voiceprints is possible, Old-school encryption too, and is already being done by commercially available The obvious answer to the problem was to encrypt the voice software from companies such as Nuance and Auraya. recording. A system would store an encrypted version of The problem is making the voiceprints themselves secure your voice, identifying you without actually having access and difficult to steal, says Bhiksha Raj, associate professor to your voice. Each system would have its own encrypted and non-tenured faculty chair at the School of Computer version, impossible to connect with each other or with your Science’s Language Technologies Institute. But Raj and his original voiceprint. colleagues may have found the way forward. To accomplish this, Raj and his colleagues had to solve two “Where (the work) really began was when we realized problems. The first was authentication itself. that every time we use our voice to authenticate “Speech is a noisy signal,” says Shanatu Rane, a principal (ourselves), we put ourselves at risk,” Raj explains. research scientist at Mitsubishi Electric Research “Your voice is supposed to be a viable biometric, but Laboratories, who collaborated with Raj on his early once you’ve given it away it becomes just another bit work. “If you say something now, and then say it again five of data out there.” minutes later, the two signals are not going to be identical.” Just as a stolen password can leave your email or financial There’s simply no way to make a person’s voice input data wide open, a cracker could potentially steal your completely stable in the same way as a typed password. voiceprint from a database and take over part of your To solve this issue, the voice authentication process life, at least as effectively as if he’d stolen your social employed by Raj’s team used Gaussian mixture models, security number. or GMMs. GMMs are a way to statistically match up a The question that stumped Raj and his colleagues was: given pattern to a standard sample. In this case, Raj says, Could they somehow get “Siri” to respond to voice the researchers used the parameters of the GMMs they commands without sending the actual voiceprint over calculated from individual voice recordings to represent the network and into the cloud, where it would be the actual recordings. Using GMMs, their system achieved vulnerable to theft? excellent results both in terms of recall (the ability to recognize a matching voice sample) and precision (the ability to avoid accepting a nonmatching sample).

10 the link. 11 In a paper written with Jose Portelo of INESC-ID/IST, INESC-ID/IST, of Portelo Jose with written a paper In could system SBE an that showed colleagues and Raj of authentication in voice precision recall and deliver encryption-based of that under 93 percent—slightly over this improved then, they Since have authentication. further. performance the same less or more actually achieve can “You a conventional with get you that performance error,” a percent a fraction(encrypted)with of system, precision recall and achieved already They’ve says. Raj suggests that level error—a 0.18 percent than less with comparing for system identification a workable is SBE the cloud. through information voiceprint the says research, in the SBE involved who not Rane, was “much are reporting are colleagues his and Raj results in accuracy takea small hit encryption. than “You faster” says. he in speed,” increase a very for large in return be can SBE before be need to resolved questions A few they need decide to Importantly, says. Raj deployed, the user’s performed on is the conversion of much how of more The in the system. done is much how device and device, theeasier in the end-user’s done the conversion device simply use to that to intruder an be for would it if example, purposes—for her everything or reset his for that the more hand, On the other phone. a smart lost you the of mercy the user at is the more in the system, resides securitygoodwill. and system’s that would be authentication voice of levels different Possibly, of conversion. formulations best served different by voice- decided, says, Raj are theseOnce issues feasible “computationally are passwords authentication today.” be can implemented and plausible, and science both of writer is a freelance Chiacchia —Ken Pittsburgh the joined recently He fact. science and fiction writer. science as senior Center Supercomputing

aking a trick from text passwords passwords text from trick a aking reconstruct the original voiceprints. the original voiceprints. reconstruct precision without any way for the hashes to be to used theto hashes for way any without precision voiceprint’s hashes in a way that maintains recall and recall and maintains that in a way hashes voiceprint’s given recording can then be compared with the original with then be can compared recording given voice signals into simplified hashes. The hashes of a hashes The hashes. simplified signals into voice embeddings,” or SBEs, to convert key features of complex complex of features key convert to SBEs, or embeddings,” Raj and his colleagues developed “secure binary “secure developed colleagues his and Raj transmit quickly. quickly. transmit as secure as encrypted data, but far smaller, and easy to and smaller, far secureas encryptedas but data, intercepting the hashed password to decode it. Hashes are are decode Hashes to it. password the hashed intercepting possesses and uses. It’s almost impossible for someone someone for impossible almost possesses uses. It’s and for your password” that only one particular one system only that password” your for “hash”—essentially, a mathematically devised “password devised “password a mathematically “hash”—essentially, to the system. Instead, it uses the password to generate a generate to uses the password it Instead, the system. to password, their computer does not transmit the password the password transmit doesnot their computer password, Most laypeople don’t realize that when they type that realize in a don’t laypeople Most more like text-only passwords. passwords. text-only like more recording, using a 1024-bit security key. It took more more took It security key. a 1024-bit using recording, behave voiceprints make to was ironically, answer, The encrypting, exchanging and decoding a 4.4-second voice decoding voice a 4.4-second and encrypting, exchanging T Computadores Investigação e Desenvolvimento tried Desenvolvimento e Investigação Computadores encrypted of data? the limitations from Portugal’s Instituto de Engenharia de Sistemas e de Sistemas Engenharia de Instituto Portugal’s from around get could they how work—so to going wasn’t That CMU graduate student Manas Pathak and colleagues and Pathak Manas student CMU graduate 14 hours. than becomes time-prohibitive. In one experiment, Raj, then- Raj, experiment, one In time-prohibitive. becomes moving it back and forth from machine to machine machine to machine from forth and back it moving encrypting them generates huge amounts of data. Just Just data. of amounts huge encrypting them generates Unlike text passwords, voice samples are complex, and and complex, are samples voice passwords, text Unlike thousand-fold.” thousand-fold.” for storage, communication and computation increases a increases computation and communication storage, for “When you encrypt data, the size of the data increases,” increases,” the data encrypt the size of data, you “When the overhead processing, encrypted-domain for suitable have plenty of time to burn. time to of plenty have considered typically keys the thousand-bit “With Rane. says and decrypting voice samples is straightforward—if you you straightforward—if is samples decryptingand voice The second problem was more of an obstacle. Encryptingobstacle. an of more was problem The second on campus Is the end near? Daniel H. Wilson’s “Robopocalypse” plots a robot uprising in the near future, but (so far) the science is still fiction

By Meghan Holohan an army of metal men to build his armor. Fictional robots It starts out slowly—so slowly people don’t realize what’s fulfill different roles in our imaginations. They have to solve happening. One robot malfunctions, attacking a human. problems and move in the physical world, like humans. Soon more robots fail, turning on humans, hunting them They’re helpers such as RoboCop, or Rosie, the Jetsons’ down and slaughtering them. Some of the survivors are maid, and they’re killers like the Cybermen of “Doctor forced into labor camps, where they undergo bizarre Who” or the Cylons of “Battlestar Galactica.” cyborg experiments. Others fight the robots. But Mason says robots do possess an essential quality that Daniel Wilson’s book “Robopocalypse” paints a bleak makes them perfect villains: “If you want to write a scary picture of the near future when robots work together to book … you need a smart enemy.” Robots are certainly crush humans and take over the world. Wilson’s books, smart, and getting smarter. The most interesting part about which include “How to Survive a Robot Uprising: Tips Wilson’s book—and what makes it different from many on Defending Yourself Against the Coming Rebellion,” other sci-fi books—is that Wilson makes a distinction focus on the dark side of technology—what happens between mere intelligence and actual human consciousness, when good robots go bad, what occurs when we become Mason says. trapped by our own creations. But he notes that according to Moravec’s Paradox—named His work, like that of so many other science fiction for Hans Moravec, an adjunct faculty member at the RI— writers, speaks to our fear of technology, and that fear it’s easier to create a fast-thinking robot than one that can resonates with readers—including Steven Spielberg, master complex locomotion. (Mason jokes that “maybe who’s currently working on the movie version of there are already evil robots out there who just have found “Robopocalypse.” Anne Hathaway and Chris Hemsworth [human] minions to carry out their plans.”) are slated to star in the film, which is tentatively scheduled Put another way, machines may be able to process for 2014 release. (Recent reports in Hollywood trade information quicker or better than humans, but it’s harder papers suggest the film is being delayed.) to train a robot to fold a towel than it is to train a robot to In “Robopocalypse,” a highly intelligent robot named think through chess moves like a grandmaster. That’s one Archos infects the world’s other robots—companions, of the ways that Wilson’s book differs from many other domestic helpers, smart homes, boxy drones, sci-fi adventures. Wilson, an SCS alumnus, doesn’t create manufacturing robots and cars—with a virus that causes a dystopia where the evil robots are humanoids, like those them to seek out humans and kill them. Many of these depicted in “Battlestar Galactica” or Philip K. Dick’s “Do robots resemble existing robots in our real-life world; the Androids Dream of Electric Sheep?” Instead, in Wilson’s self-driving cars bear an eerie resemblance to driverless world, boxy hunks of metal perched on insect-like legs vehicles developed at CMU, Google and elsewhere, while lead the bloody revolution, while smart cars weed out the the humanoid robots sound like a more advanced version human population. of the Wabian robots at Japan’s Waseda University. “There In other words, Wilson’s robots are only a bit more advanced isn’t anything inherently scary about robots,” says Matt than our current technology, making it seem as if there Mason, professor of computer science and director of could be monsters in labs and even in our homes. It makes CMU’s Robotics Institute. Sometimes they’re bad, other a robot like Archos—seemingly limited in his mobility— times they’re good. a more realistic threat than an army of murderous humanoid Indeed, asks Mason, “Isn’t the reason that we’re interested robots yet to be devised. in robots because they’re so much like us?” Are robots currently as smart as those depicted in Wilson’s Robots have played important roles in literature, TV books? Self-driving cars and automated drones are certainly and movies for generations—centuries, if you count sophisticated, says Chris Atkeson, a professor of robotics Homer’s description in the “Iliad” of Hephaestus using and human-computer interaction at CMU. He also served

12 the link. 13

What makes so many people so terrified of robots? “I am“I robots? so people of terrified so makes many What that noting Atkeson, says psychoanalyst,” play to going up critical A growing part of children. like “act robots and parents rejecting your is independent becoming and stronger are Robots own thing.” your doing and off going the this idea with Combine humans. than smarter and robots— mistreat would believe humans many that easy to it’s citizens—and them second-class making “grow robots as revolution horrible a resulting imagine seek vengeance. and “parents” reject theirhuman up,” Pittsburgh’s at takes place book, “Amped,” latest Wilson’s what explores School and High Allderdice Taylor their from inseparable become whenpeople happens have thecharacters of some “Amped,” In technology. help to intended originally were that neural implants these neural implants But handicaps. overcome people humans. other to smarter—superior people making are a civil-rights a story of techno-thriller, another “It’s disabilities with when people sparked is that movement says. Wilson smarter,” become to useneural implants robots the us fighting between the difference really “It’s the robots.” becoming us and “Robopocalypse” to a sequel on work at now is Wilson the technology and exploring continue to plans and more pro-technology—the am “I us. over holds it power technology we the more “but says, he the better,” power it.” depend on lives our the more have, . Wilson Daniel H. animated robot who saves the world. the world. who saves robot animated robots and one of its most popular heroes is Astro Boy, an Boy, Astro is heroes popular most its of one and robots varies by country. In Japan, for instance, people embrace embrace people instance, for Japan, In country. by varies becoming too powerful.” But he adds that fear of robots robots of fear that adds he But too powerful.” becoming “We’re afraid of machines because we’re afraid of them of afraid because we’re machines of afraid “We’re and more of an impact on people’s lives,” Wilson says. says. Wilson lives,” people’s on impact an of more and following year. “Technology and robotics are having more more having are robotics and “Technology year. following in 2005 and won Wired magazine’s “Rave” award the award “Rave” magazine’s Wired won in 2005 and “How to Survive a Robot Uprising,” which was released which was Uprising,” Survive a Robot to “How he began writing the semi-serious became writing began bookhe that and robotics. He was still in CMU’s Ph.D. program when program Ph.D. still was in CMU’s He robotics. and resonated until he started studying real computer science computer real startedstudying he until resonated But Wilson didn’t really feel that his science his fiction that feel really didn’t Wilson But excited.” excited.” was such a thing as (artificial intelligence), I got really got I (artificial as a thing intelligence), such was computers to do creative stuff. Thenwhen I learnedthere stuff. do creative to computers gave me a real creative release … I was able to program program to able I was … release creative a real me gave “Ultimately, I learned how to program as a kid and that that a kidas and program to I learned how “Ultimately, fiction “like a lot of scientists do,” he says in an interview. interview. an in he says do,” of scientists a lot “like fiction education at CMU. Growing up, he enjoyed science enjoyed he up, Growing CMU. at education Wilson says his fiction is informed and inspired by his inspired and is informed fiction his says Wilson haven’t achieved that yet.” yet.” that achieved haven’t we have now, in that they are thinking like people. And we And thinking people. like they are that in now, have we greater cognitive capacity in (Wilson’s) robots than what what than robots in (Wilson’s) capacity cognitive greater subtle thinking.” But he adds a caveat: “I think there is a is think there “I a caveat: adds he But thinking.” subtle decisions on their own,” Atkeson says. “That takes pretty pretty “That takes says. Atkeson their own,” on decisions drive themselves and those cars have to make tough tough make to have those cars and themselves drive as Wilson’s Ph.D. advisor at CMU. “Google has cars that that cars has “Google CMU. at advisor Ph.D. Wilson’s as on campus in the loop: john c. reynolds What made you decide to retire? I’m suffering from some health problems and I wasn’t sure I would be able to teach in the spring. But now that I’ve found out how much work it is to retire—you’d think that if there was one thing that would be made simple, it would be retirement—it’s a little scary!

Where did you grow up, and how did you get interested in science? I grew up in the Chicago suburbs, in Glen Ellyn, Ill. My mother had been a schoolteacher and my father was in advertising. It was apparent by the time I got to school that my first love was music, and I thought I might become a musician—I played piano and trumpet. But somehow, science became more and more interesting. I wound up being one of the winners of the annual Westinghouse Science Talent Search, went to Purdue and did my graduate work at Harvard in theoretical physics.

One of your teachers at Purdue was Alan Perlis? Yes—he left Purdue for Carnegie Tech in 1956, at the same time I graduated. While he was there he taught the one and only course in computing. It was half a semester on how to build them, and half on how to program them. One day, he walked into class and said, “I’m going to show you how to build a divider.” He started sketching this long equation on the board. Finally, one student said, “Professor, that won’t work John C. Reynolds has been a professor of computer science with a negative quotient.” So he made some changes, at Carnegie Mellon University since 1986. He retired from and another student said, “Professor, now it won’t work with a positive quotient.” He made some more changes, active teaching Jan. 1. A graduate of Purdue University, and someone said, “It won’t work with a positive Reynolds earned his Ph.D. in theoretical physics from divisor!” Perlis said, “Well, that just goes to prove that Harvard University. you cannot design a divider on a Monday morning. Class dismissed!” His main research interests have been the design and definition of programming languages and the specification What was the state-of-the-art then? of program behavior. Strictly machine language. This was before even A past editor of the Communications of the ACM and the Fortran. Perlis at that time was working on one of the Journal of the ACM, Reynolds was appointed a fellow of the first compilers. ACM in 2001 and won the ACM SIGPLAN Programming Language Achievement Award in 2003. Reynolds was When you got to Harvard, was it more up-to-date honored with the Lovelace Medal of the British Computer than Purdue? Society in 2010. Harvard at the time was under the spell of Howard Aiken, which means its computer system was He spoke to Link Editor Jason Togyer. completely obsolete. Fortunately, IBM had set up an installation at MIT that could be used by a consortium of other universities, and I had access to that.

1414 the link. the link. 15 15

irard’s, has beenhas Girard’s, olymorphism—the ability to handle many different data different many handle to ability olymorphism—the out in computer science? computer in out thatis One said! be to need that things many so are There mastercompletely to have they grad feel students many apartQuite thefrom anything. do can they before field their ofamount shouldn’tIt stultifying. it’s take, would that time to goal your be should It learnto everything. goal your be learn.to have people other that things discover retirement? in pursuinghobbies any you Are problem a it’s But time. to time from piano play still I ifbecause of level some reach you thenand technique ofbit a It’s back. it get to painful really it’s go, it let you startto like would I again. but playing struggle, uphill an dance and musical to going love also I and wife my Both catching job good performances,pretty a doing I’m and reading. my with up useful in reducing higher-order functions to simplerto functions higher-order reducing in useful ofamount fair a had have Those abstractions. influence. of extension an is logic—which separation third, And the logic— Hoare in use normalyou that calculus predicate determineto of correctness the program.a starting just student a for have you would advice What credited with leading to most ofmost to leading with credited modernthe computer agreethat? with you Would languages. influence on much that all had I’ve sure not I’m Oh, ofidea The polymorphism. waspolymorphism really ifthat 1967, in back Strachey Christopher by conceived you sortinga “something,” have any program for work will it and stillis Britain that reason the he’s and polymorphic, is it there film, Bond James latest the In field. the in leader the ofkind some displays computer a which at point a is a “geek” the and larger, times three graphgets suddenly that polymorphic!” gone it’s heavens, “Good says, character Solaughing! out burst I and wife my that, saw we When influence culture? pop on tremendous a had I’ve maybe influential? been has work your think you do Where work my also but extent, some to Polymorphism, oflimits the interpreters—defining definitional on a easy-to-understandan writing programmingby language beenhas which defunctionalization, interpreter—and which doesn’t sound that unusual now, but was unheard of was unheard but now, unusual that sound doesn’t which usefulbe could it that was show to contribution My then. discovery real my and parameters, as around types pass to language polymorphic a type-check could wasyou that errors different between catching words, other in statically, oftypes data. P become programmingsame the with types language—has with along work, your and ubiquitous,

ow did you you did How CMU? Harvard? irard, you’re credited with credited you’re Girard, ves ves Jean-Yves is a first-rate logician. He took that concept and logician. concept that took He first-rate a is Jean-Yves afrom have could I than deeply more much it explained program every that you proved He standpoint. theoretical some but finish, always will calculus the in write could inventing polymorphic lambda calculus. calculus. lambda polymorphic inventing timesame the at thing same the on working be to come places? different in finish— to time long incredibly an programs such take will Y Jean- with Along The fact that I could design the rules ofrules the design could I that fact The languagethe very toeasy were that principles essential few a on based terriblea learn.and implement have I think I back, Looking about me excited that thing big the probably and memory, you which by principles general were wasmath there that of, nature ad-hoc the have didn’t it and things, derive could history. say, What eventually brought you to you brought eventually What ofsemantics the on working to you attracted What programminglanguages? From 1961 to 1970 I worked at Argonne National Argonne at worked I 1970 to 1961 From Argonne at worked had I gradin school, While Laboratory. home, wasit because close summers, the during my to would they because Harvard after Argonne to went I and move or physics in stay either to me with happy been have in very interested gotten had I time, that By computing. to ofone and programmingand compilers languages, the compilera was designed I things first compiler—actually, a ofkinds the because COGENT, called compilers for suitable not programmingwere then had we that languages manipulation.symbolic for highthe and reputation the were me attracted what Initially, withpeople have to have You here. saw I research quality theon but hand, one the on goals common share you whom andskills different bring who people have to have you other of many CMU, At the other. each complement and talents programmingfunctional on working languages, were people programming imperative on work to tended I while thatin stronger research my made that’s think I languages. myselflimited haven’t I other. the or side one to Where was your first after stop your was Where Well, my Ph.D. was called “Surface Properties was “Surface called of Ph.D. my Well, Nuclear wasit number-crunching big a actually but Matter,” physics program. wasI think I theoretical in person first the thatand thesis, his do to computer a used who Harvard at should they than more faculty the impressed probably uninteresting “an as it describe I impressed. been have ofcomputation unimportantan bada in quantity approximation”! ow does one go from theoretical physics to computing? to physics theoretical goone does from How feature

theSmashing stereotypes

16 the link. 17

CS and men) into men) and subassembly to loosen partsSoon, to and boards, it. subassembly By the the floor. on piles in neat stacked are cables they features listing are the girls the night, end of they things didn’t and their computers, liked about one things,” USB needs more computer (“Our like. the of top on was supply power our “and girl says, open.”) get to hard insanely was it and motherboard rawing more women women more Drawing ( the Pittsburgh on Mondays held “TechNights,” experience the women let theseyoung campus, computer in this case) of (literally, bolts” and “nuts multifaceted ongoing, an a part of It’s science. to booststagnant the Women@SCS by effort in computer seeking careers women of percentages to and technology, information science and the field beyond entering the men pooldiversify of hackers. and male gamers Asian and white male students of number increasing fact, an In outreach Women@SCS help to volunteer now CMU where “roadshows,” including programs, Pittsburgh-area visit science students computer talk them and to high-school students middle- and the of many “And robotics. and computing about “They says. Frieze minorities,” are involved men be not to like it’s what understand it’—they ‘get in a field.” represented involved also are volunteers Several Women@SCS helping called ScottyLabs that’s organization in an undergrads, CMU among culture “maker” spread is which TartanHacks, called effort related its and “hackathons.” of the appeal designed broaden to ethical, reasons—moral, abstract are There and women encourage to wanting cultural—for careers pursue to groups under-represented other math, and engineering technology, in science, practical reasons are there But fields. “STEM” or the U.S. from 2011 data to According well. as in STEM people Commerce, of Department hour per more 21 percent about earn fields

ogyer T

and other under-represented groups alone less feel under-represented other and

who aren’t white male gamers—and male womenmaking white aren’t who Computer science’s future depends on attracting people attracting on depends future science’s Computer One girl is jumping up and down on a stubborn a stubborn on down and up jumping girl is One supply. They aren’t being dainty and delicate. delicate. and dainty being aren’t They supply. drive, memory chips, processor, fan and power power and fan processor, memory chips, drive, Each girl is expected to find the hard drive, CD drive, Each expected girl is hard find the to like a regular classroom, and it’s not meant to be.” to meant not it’s and classroom, a regular like cautiously from a corner. “As you can tell, this is not this not is tell, can you “As a corner. from cautiously Carol Frieze, director of Women@SCS, watching watching Women@SCS, of director Frieze, Carol we complain because they won’t take risks,” says says take risks,” because they won’t complain we tell them ‘be careful, don’t hurt yourself,’ and then and yourself,’ hurt don’t them ‘betell careful, “We tend not to let girls do things like this, or we we this, or like do things let girls to not tend “We screwdrivers, wrenches and pliers. pliers. and wrenches screwdrivers, computers—first with pryingthen with fingers, computers—first safety rules, the girls are tearing apart their apart tearing rules,safety the are girls engineer a computer.” After Hildebrandt reviews Hildebrandt After a computer.” engineer “Tonight,” she adds, “you are going to reverse- to going are “you adds, she “Tonight,” changing the ingredients. changing something and figured out how to make it better by it better make to how out figured and something Engineering,’” Hildebrandt says. “You took apart apart took “You says. Hildebrandt Engineering,’” “You just did something called did something ‘Reverse just “You like?” Really nasty, the say. girls like?” Really nasty, too much! And what would thosecookies would taste what And too much! “Right!” Hildebrandt says. “A half-cup of salt is way way is salt of half-cup “A says. Hildebrandt “Right!” gets the right answer—too much salt. much answer—too the right gets chips. Another says “too much butter.” Finally, one one Finally, butter.” much “too says Another chips. Hands shoot up. One girl says too many chocolate chocolate toomany girl says One up. shoot Hands is the wrong amount. Do you know which one?” know Do you amount. the wrong is quiz for you,” she says. “One of these ingredients theseingredients of “One says. she you,” for quiz the screen listing cookie ingredients. “I have a pop a have “I cookie ingredients. the screen listing undergraduate student organizer, puts a slide on on a slide puts organizer, student undergraduate Hildebrandt, a senior computer science major and and science major computer a senior Hildebrandt, by CMU’s Women@SCS program. Natalie Natalie program. Women@SCS CMU’s by nights, or “TechNights,” for girls presented presented girls for “TechNights,” or nights, This is one of the weekly Creative Technology weekly Creative of the one is This about to give its life for science. for life its give to about Computer Science’s IT team. Each computer is Each computer IT team. Science’s Computer old desktop computer, donated by the School of by donated computer, desktop old the floor. In the middle of each group there’s an there’s of group In each middlethe the floor. middle-school girls are sitting cross-legged on on cross-legged sitting middle-school are girls Room in the Hillman Center. Groups of squirming of Groups Center. Room in the Hillman It’s a Monday night at the Raj Reddy Conference the Reddy Raj Conference at night a Monday It’s y Jason Jason By FEATURE

than people in non-STEM jobs. Women in STEM jobs earn Allan Fisher (CS’81, ’85) and Jane Margolis, a senior 33 percent more than women in comparable non-STEM researcher at UCLA, published “Unlocking the jobs, the same data indicate. In February, the National Clubhouse,” a much-discussed book about the barriers Center for Women & Information Technology announced faced by women pursuing CS degrees and ways to that 20 universities and 14 companies had signed onto overcome them. (Much of their research was influenced by “Pacesetters,” a two-year program that encourages senior work being done at CMU, though Blum, Frieze and others executives to recruit women for technical careers from say “Clubhouse” does not necessarily reflect current or past previously unused talent pools, and develop strategies to SCS practice.) retain women who are at risk of leaving computing and IT But the short-lived increase in women applicants couldn’t careers. CMU has been a member of NCWIT’s Academic be sustained; along with the end of the NSF outreach grant Alliance since 2004 and is a partner in Pacesetters. came the early 2000s dot-com bust, and the numbers Having a more diverse pool of people working in the field declined. Today, the percentage of female computer will pay off in innovations, argues Lenore Blum, CMU science undergrads at CMU is about 25 percent. Yet distinguished career professor of computer science and Carnegie Mellon is still doing better than peer institutions, founding director of Women@SCS. She currently serves many of which have percentages near 10 or 15 percent. as the group’s faculty advisor. “When you increase the variety of the people involved, they start paying attention to more different kinds of things,” Blum says, “and there’s ‘Paint it pink’ a bigger potential for some big disruptive technology Throughout the U.S. and Europe, women remain a distinct breakthroughs.” minority in computer science programs, and faculty such as Blum say some universities took the wrong lessons from the work done by Fisher and Margolis: They watered Networking, leadership, outreach down curricula in attempts to get women to enroll in Women@SCS is not a club, department or office. It’s an computer science programs. Blum calls it the “paint it advisory council and a collection of student-led working pink” mentality—trying to attract women by emphasizing groups (with assistance from faculty and staff) that focus the uses of computer applications rather than the on sharing information about women in computer science fundamentals of programming, systems and operation. both on campus and in the wider community; promoting “If you want to know how to create computer programs opportunities for women to contribute research and rather than just use them, that’s not good,” she says. ideas; and connecting female students to mentors. The Blum has been working to increase the numbers of program has its roots in the late 1990s, when the School of women in science and technology careers since she was a Computer Science, under then-Dean Raj Reddy, made a faculty member at California’s Mills College in the early concerted effort to encourage female high school students 1970s, where she helped launch the first computer science to apply and create a support structure to keep those department at an all-women’s college, and served as co- students enrolled. It was launched with money provided director of Mills’ “Expanding Your Horizons” math and by CMU President Jared Cohon and received continued science conferences for high school girls. “In some ways, funding from SCS deans Raj Reddy, Jim Morris (CS’63) progress hasn’t been as fast as I would have hoped,” and Randy Bryant, Blum says. Blum says. In fact, she says CMU may be unique in that its computer The reasons for the lack of progress are complicated and science outreach to under-represented groups has a have social and political overtones. The American feminist dedicated, continual funding stream. “In other places, with movement of the 1970s was followed by a backlash in other programs, they’re funded by grants—and when the the 1980s from people who considered it a threat to grants go away, the programs go away,” Blum says. traditional nuclear families. Dwindling budgets for social At the same time Women@SCS was created, CMU programs meant that funding for some early efforts to also received a National Science Foundation grant to encourage women to pursue science and technology (such fund outreach to high school computer science teachers. as federal resources made available through the Women’s In a single year, the percentage of women entering the Educational Equity Act) faded away. SCS undergraduate program shot from single digits to near 40 percent. In 2002, then-SCS Associate Dean

18 the link. 19 Consequently, most middle- and high-school-age students students high-school-age middle- and most Consequently, science unless computer about learn much to unlikely are the of many where the classroom, outside it they explore heroes—Facebook’s or pioneers tech as up held people Jobs Steve Apple’s Bill Gates, Microsoft’s Zuckerberg, Mark of visible males.lack The white Wozniak—are Steve and of people and women for models science role computer of director Goldfein, Jocelyn problem, a serious is color “I’ve Post. Huffington The told Facebook, for engineering basically to believe this a self-fulfilling is prophecy,” come more aren’t “The reason there the website. told Goldfein very many aren’t there science that is in computer women a computer look into You science. in computer women this think, ‘Oh, and see men and mostly science classroom has class that a find go to going I’m me. for not is classroom me.’” like look more that people more men than more have fields all true “STEM” that It’s A Gender in STEM: “Women to According women. Commerce the U.S. by a 2011 report in Innovation,” Gap the of 48 percent up make women although Department, jobs. STEM of 24 percent only they hold workforce, U.S. science particularly is acute in computer the problem But Between 2009, direction. 2000 and in the wrong going and as engineering such fields in STEM women of the number whilein slightly, up sciences went life and physical and the technology, information and math science, computer 3 percent. fell women by jobs held of number classes fading, too fading, classes CS

says. “A lot of kids don’t realize that it’s even an option.” option.” even an it’s that realize kids don’t of lot “A says. out there was such a thing as computer science,” Quispe science,” computer as a thing such was there out science and technology. “It wasn’t until then that I found I found then that until wasn’t “It technology. science and High School, which offers accelerated programs in math, math, in programs accelerated School, which offers High was fortunate enough to attend New York City’s Stuyvesant Stuyvesant City’s York New attend to enough fortunate was senior majoring in computer science. The Queens native native The Queens science. in computer majoring senior know that math was useful,” jokes Amy Quispe, a CMUQuispe, a Amy jokes useful,” was math that know “When I was in sixth grade, I was good at math, but I didn’t I didn’t but math, sixth in good I was I was grade, at “When viewed as “menial” work. viewed “menial” as that she got the feeling that computer programming was programming computer that the feeling got she that teacher “looked at me like I was insane,” she says, adding adding says, she insane,” I was like me at “looked teacher the computing industry as a career path,” Davis says. One says. Davis path,” industrya career as the computing remember expressing my desire to pursue something in something pursue to desire my expressing remember She knew of only one girl that took the class. “I also “I the class. took girl that one only knew of She if computer science were a vocational course, like typing. typing. like course, a vocational science were if computer listed computing classes under “business technology”—as “business under classes computing listed CS major, says her high school in southern Maryland high her in southern school says CS major, programs being de-emphasized. Elizabeth Davis, a junior Elizabeth a junior Davis, beingde-emphasized. programs art, music, social studies and foreign languages, it’s among among it’s languages, foreign social and art, music, studies Computer science is treated as an elective, and along with along and elective, an as science treated is Computer that aren’t subject to mandatory standardized testing. testing. standardized mandatory to subject aren’t that math and reading scores to the exclusion of programs programs of the exclusion to scores reading and math facing budgetary pressures have focused on improving focused improving on have budgetary pressures facing (The Link, Spring 2011) pointed out, public school systems systems school public out, 2011) pointed Spring (The Link, science these days. As the 2010 study “Running on Empty” Empty” on “Running the 2010 study As science these days. or female—are getting any formal education in computer in computer education formal any getting female—are or igh-school High-school high school students—male American few matter, that For FEATURE

And the problem in computer science is also odd because The stereotype continued long after it had any basis in it’s among the youngest of the scientific disciplines, and reality, Ensmenger says. “Take other professions that we because there wasn’t always such a big gender gap. Women recognize as taking a lot of time commitment—medical such as Ada Lovelace and Grace Hopper are among students, for instance, work long hours and pride computing’s earliest pioneers. All of the programmers themselves on hard work, but they don’t think being scruffy of the original ENIAC at the University of Pennsylvania or nerdy is one of the required attributes for proving that were women. you’re smart or competent,” he says. Blum notes that those women made in-roads during World War II; following the war, there was a concerted effort to force women out so that male veterans could Persevering in a ‘very male culture’ return to those jobs. Nathan Ensmenger, a professor of That image of the stereotypical computer nerd, reinforced informatics at Indiana University who has researched by movies and T.V., has attracted like-minded people while historical reasons for the gender imbalance, says a variety simultaneously discouraging women and many men from of other factors also combined to discourage women from computer science careers. Gabriela Marcu, now a Ph.D. careers in computer science. student in CMU’s Human-Computer Interaction Institute, has first-hand experience with the stereotypes from high school and college. With encouragement from her mother The roots of disparity (an engineer), Marcu took AP computer science in high school, “but almost nobody else took the class,” she says. Many early computer programmers were recruited into “At first, I thought they were a bunch of nerds, and I didn’t “data processing” departments from secretarial pools, want anything to do with it.” which were overwhelmingly comprised of women, Ensmenger says. As computers became more important, Later, when she got to college, she can remember the programming jobs grew in status and salary, attracting “atmosphere of male competitiveness.” “Some of these more interest from men. In attempts to professionalize guys had been programming since they were five,” Marcu the field, businesses began recruiting people who had says. “Every class, every assignment, they had to prove four-year college degrees in science and engineering; in the everything they knew.” Her early interactions with her 1960s, that favored men. And some male managers simply mostly male classmates filled her with apprehension. “As a didn’t want to trust their increasingly crucial computer new student, that made an impact,” she says. “I felt like, ‘No operations to women, so they didn’t hire them. “There is in matter how smart I am or how hard I work, I’ll never catch computer science a kind of sexism and misogyny that’s not up, they have so much more knowledge.’” deeply hidden under the surface,” says Ensmenger, who has With encouragement from faculty, Marcu persevered. expanded his research into a book about the phenomenon “I found out that computer science isn’t just about called “The Computer Boys Take Over” (MIT Press). programming, and those people who are super-duper There were other factors as well. In the ’60s and ’70s, technically and want to impress you with their technical programmers often were only allowed access to computers skills generally have very limited knowledge in a certain after the day’s data processing was done—in the late area, and don’t have great social skills,” she says. Still, it took evening and early morning. But because some colleges time for Marcu to shake the feeling that she was an outsider and corporations didn’t allow women on the premises in the world of computer science. “It was a very male overnight by themselves (supposedly for “safety” reasons), culture,” she says. “It’s not that I doubted my ability, but I the programmers were mainly young, single men. Working felt like maybe it wasn’t the place for me.” in isolated computer rooms away from other professionals, That feeling of being an “outsider” can become, as programmers cultivated images as rebels—unshaven, Goldfein put it, a self-fulfilling prophecy. “Numbers make unshowered, uncouth, anti-social—spending time alone a difference—people need to be able to see themselves in in dimly lit labs, fueled by junk food and rejecting anyone the field,” Frieze says. The ratio of male to female computer unwilling to live under those conditions as not dedicated to science faculty also has a “subtle, not deliberate” effect, the craft. By the 1980s, the stereotype would be known as she argues. “Some women can go through an entire CS the “computer nerd” or “hacker.” undergraduate program without ever having a female faculty member” as an instructor, Frieze says.

20 the link. 21

o special treatment special No do you find out about things?” Blum says. “If you’re the you’re “If says. things?”Blum about out find do you who do class, in your students eight of out woman only her male colleagues would be excited about new findings, new findings, about be excited male would colleagues her know didn’t co-workers female that gossip theories and excluding consciously they were that wasn’t It about. when they socialized that the on just together it’s women; information. they shared weekends, how students, eight of out woman the only you’re “If of advantage … The a problem? on help call for up you you a network, part of you’re that is being in the majority connections.” have the gender to inured become they’ve say students Some a conceptual on me bothers “It adapted. and imbalance basis,” a day-to-day on bother me really doesn’t it but level, I’ve and up male friends growing of a lot had “I says. Clute pay people making in and jumping a problem had never SCS provides feels she importantly, More me.” to attention assert to her herself. allows that environment a nurturing used completely just I’m this point beat might “It adds: She I look around home go I When guys. by beto surrounded here?’” women so many there are ‘Why say, and have suggested that women are more interested in interested more are women that suggested have whiledesigns, men interface and applications Blum: SCS’s has beenoccasionally against a backlash There began SCS first When gap. gender the to address efforts counselors admissions asked Reddy efforts, outreach its In undergraduates. accepting for the criteria broaden to them asked he high to academic performance, addition Because potential. leadership with students look for to theirown coding skills who developed had on students no that told were students be prospective to male, tended “People experiencenecessary. was programming prior are these women of ‘All complaining, calling and were says. Blum son,’” my taking of the place to not SCS for she says, important, it was why That’s supposed curriculum to its appeal to down water instance, for researchers, Some differences. gender

As computers became more important in the 1960s and ’70s, programming ’70s, programming in the1960s and important became more computers As some And men. from interest more attracting salary, and in status jobs grew crucial trust computer to their increasingly want didn’t simply malemanagers them. hire so they didn’t women, to operations

at Berkeley, Blum remembers how on Monday mornings, mornings, Monday on how remembers Blum Berkeley, at director of the Mathematical Sciences Research Institute Institute SciencesResearch the Mathematical of director available to those in the minority. When she was deputy deputy was she When those to in the minority. available the majority in a group offers professional advantages not not advantages professional offers group in a the majority ways difficult to quantify. As Blum points out, in being points Blum As quantify. to difficult ways experiences is important to professional development in development professional to experiences important is Providing a network of people who share common common who share people of a network Providing been exposed to a program like TechNights when she TechNights been like exposed a program to logic.’” it’s magic, came to CMU as a cognitive science major, wishes she had she wishes science major, cognitive a CMU as to came not it’s ‘Hey, I said, too, do this, I could that out figured committee and also volunteers at TechNights. Clute, who Clute, TechNights. at also volunteers and committee the impression under always was in high “I was school. “Once I says. Clute magic,” of off ran computers that from Concord, Mass., who chairs Women@SCS’s outreach outreach Women@SCS’s who chairs Mass., Concord, from the same classes before,” says Madeleine Clute, a CS junior a CS junior Clute, Madeleine says before,” classes the same interests and are in the same classes as me, or have taken have or me, as classes in the same are and interests “It’s nice to have female friends who share many of my my of many friends female who share have nice to “It’s Hopper Celebration of Women in Computing. Women of Celebration Hopper who want to attend conferences such as the annual Grace Grace the annual as such conferences attend to who want frankly); and provides financial support for students students for support financial provides frankly); and as “no faculty allowed,” so that students discuss issues discuss issues students so that faculty allowed,” “no as social hours (some, for both men and women, are billed are women, both and men for social (some, hours Sister/Big Sister” program; sponsors regularly scheduled regularly sponsors program; Sister” Sister/Big incoming undergraduate women with mentors in a “Little in a “Little mentors with women undergraduate incoming In other networking efforts, Women@SCS matches matches Women@SCS efforts, networking other In Computer Science. Computer Veloso (CS’89, ’92), CMU’s Herbert Simon Professor of of Professor Simon Herbert ’92), CMU’s (CS’89, Veloso Corp., will be the keynote speaker, along with Manuela Manuela with along will speaker, be Corp., the keynote 20. Mary Ann Davidson, chief security officer for Oracle for Oracle security chief officer 20. Mary Davidson, Ann graduate school. This year’s workshop will held Oct.be 18– workshop year’s This school. graduate science, explore research opportunities and talk about talk and about opportunities research explore science, undergraduates to work together on problems in computer in computer problems on together work to undergraduates workshop sponsored by Women@SCS that allows female female allows that Women@SCS by sponsored workshop alumni also participate in OurCS, an annual three-day three-day annual in OurCS, an alsoalumni participate collaborate with female faculty and alumni. Faculty and and Faculty alumni. faculty and female with collaborate and research events, students are encouraged to meet and meet and to encouraged are students events, research and scientists is an important goal of Women@SCS. At social At Women@SCS. goal of important an is scientists Raising the visibility of female and non-white computer computer non-white and female of theRaising visibility roviding a visible support visible a network Providing FEATURE

are more interested in coding and hardware. Blum says “proselytizing” for computer science. “Our students do a her own research, as well as 40 years’ experience as an lot of outreach,” she says. “So many of them say, ‘I never educator, disproves those theories. realized computer science was so great, and I don’t want these high school kids not thinking of it.’ They feel a duty “There are not intrinsic gender differences,” she says. to go back and share it, because they don’t want anyone “There are internal differences within genders. There is else to miss out.” a spectrum, where some men like coding, and others like applications, and some women like coding, and Getting kids excited about computer science means others like applications.” Trying to target perceived “showing off the magic,” says Kenny Joseph, a Ph.D. gender differences is “counter-productive,” Blum says. student in the Institute for Software Research who volunteers through Women@SCS to work with students Frieze, who has collaborated with Blum on research in grades six through eight at the Pittsburgh Science and into gender imbalances in computing, surveys CMU Technology Academy, a public school not far from the undergraduates on a variety of topics, and says the CMU campus. That includes demonstrating computer differences between men and women are statistically science and robotics with experiential, hands-on methods, insignificant. “We’re up against this argument that ‘Men such as the Alice software developed at CMU. are from Mars, women are from Venus.’ I’m not in that camp at all,” she says. Many perceived gender differences One bit of “magic” that never fails to thrill middle schoolers are the result of cultural and environmental conditioning, was created by Blase Ur, a Ph.D. student in computer she maintains. “The good news is that if you’re up against science, who put specially engineered Ardunio micro- culture, culture can be changed. Sometimes it’s very slow, controllers into a variety of household appliances that but it can be changed.” allow them to be operated remotely from the children’s PCs using BYOB, a custom implementation of Scratch— a programming tool developed at MIT for use by kids. Changing the culture “I’ve never seen kids so excited as when they turn on a toaster oven from across the room and melt cheese,” For volunteers in Women@SCS, one way to change Joseph says, with a grin. “When I see the kids smiling and the culture is by reaching out to younger children at engaged, I know they’re enjoying themselves.” elementary, middle and high schools in the Pittsburgh area through roadshows and TechNights. Blum calls it With kids who aren’t yet in high school, it’s a little early to

22 the link. 23

high school boys being high school boys,” she says: Loud, Loud, says: she beinghigh highboys school school boys,” video discussing games. and pranks pulling boisterous, of a group but off, me turned the truth, it you tell “To too.” off, me turned have would that actinglike women at course when Schervish math a discrete took Later, these all guys of were “there student, a pre-college CMU as being remember I just and questions, answering in the class be could good guys that remember because I didn’t shocked too.” math, at who pursued high school classmates Schervish’s of Many she chemistry, biology or into gone in science have careers she classes, science computer offered The Ellis says. School seen was It every kids signed up four year. “maybe but says, are ‘Why science), like, take (computer to strange kind of as in that?’” interested you computer that reason another story illustrates Schervish’s students women—female for be scienceplace can a difficult they’re side, one On directions. pulled in different feel femininity. of stereotypes cultural to conform to pressured the perceived to conform to pressured they’re On the other, act feminine. less and culture “nerd” women other criticize women heard has Davis instance, For of computer in so-called areas soft research pursuing for same face that don’t design. Men interface as such science, comic XKCD an to it compares Davis says. she criticism, In Randall Munroe. roboticist) (and cartoonist by strip an on working are figures stick male two panel, the first figures the stick of one math,” at suck you “Wow, equation. panel, thesecond In a mistake. when partner his makes says working on figure stick female a and figure a male stick the male math,” at suck girls “Wow, problem. the same on the exact same makes female figure when the says figure notice when escape that Things point? Munroe’s mistake. behavior female be to “typical” assumed do them are men does them. when a woman has been from discrimination felt I’ve “The only people women many how wonder “I says. Davis females,” other way.” the same in technology feel to extends feminine” “too to be not pressure The sarcastic heard have CS students Female appearance. dresses in skirtsor “girly” dressed when they’ve remarks most like T-shirts and sweats jeans, wearing of instead male students. “Sometimes, someone just wants to cut to wants just someone “Sometimes, male students. you down,” Quispe says. But the criticisms sting, the criticisms But Quispe says. down,” you

Female students feel pulled in different directions. On one side, they’re they’re side, one On directions. pulled in different feel students Female they’re On the other, femininity. of stereotypes cultural to conform to pressured act feminine. less and culture “nerd” the perceived to conform to pressured

eer pressure a factor a pressure eer she entered a co-educational setting. “Some of it was just just was it of “Some setting. a co-educational entered she School, remembers experiencing some culture shock when shock culture experiencingsome School, remembers Schervish, who was attending Pittsburgh’s all-girls Ellis Pittsburgh’s Schervish, who attending was science and math. science and CMU, too,” says Schervish, a senior majoring in computer in computer Schervish, says majoring a senior too,” CMU, find out they enjoy it,” he says. it,” they enjoy out find at now “She’s group. summer’s in that girls the other of perception that they should try it, they’re never going to to going never try they should that they’re it, perception friendsmade one with thrived and she Schervish But says. would be really good at this, but if they don’t have the have if they don’t this, but be good really would at camp,” be to nerd going was it thought “I school students. “It boils down to the fact that there are lots of kids who of lots are there the fact to that down boils “It high for program enrichment SCS summer an Leap, Joseph hopes the outreach helps chip away at stereotypes. stereotypes. at away chip helps the outreach hopes Joseph in Andrew’s participate to professor, a CMU statistics really important to change that,” he says. he that,” change to important really Mark, father, her Schervishscience. by encouraged was mediated by computers in some way, and therefore it’s it’s therefore and way, in some computers by mediated computer enjoyed she that know Schervish wouldn’t color, Ensmenger says. “There’s no job today that’s not not that’s today no job “There’s says. Ensmenger color, Margaret program, outreach a similar been for hadn’t it If pressure closes off entire career paths to young men of men young to paths career entire closes off pressure P computer ‘geek’ is seen as behaving ‘white.’” The The cultural ‘white.’” seen is behaving as ‘geek’ computer young African-American men,” he says. “To be seena as “To says. he African-American men,” young IU professor. “There’s a restricted range of masculinity for for masculinityof restricted range a “There’s IU professor. Joseph’s observation is important, says Ensmenger, the Ensmenger, says observation important, is Joseph’s friends doing that activity? No? Then I can’t do it.’” do I can’t Then No? activity? that friends doing they should,” he says. “It sometimes comes down to, ‘Are my my ‘Are to, down comes sometimes “It says. he they should,” looking at an activity like they want to try it, but aren’t sure sure aren’t try to but they like activity want it, an at looking than a racial divide. “You can kind of see of kind kids sometimes can “You a racialthan divide. noticed among his middle schoolers less of a gender divide a gender of less middle his schoolers among noticed folks. Joseph, who’s researching social networks, has has social networks, researching who’s Joseph, folks. perception that computer science is something for white white for science something is computer that perception Yet even among preteen kids, it’s hard to smash the smash to hard kids, it’s preteen even among Yet exposed to this kind of thinking,” Joseph says. says. Joseph exposed thinking,” this kind of to trying these otherwise kids who be would push to not rather than just being passive consumers. “Part of it is just just is it of “Part consumers. being passive just than rather that they have the power to invent and control technology, technology, control and invent to the power they have that problems using computational thinking, or to show kids show to thinking, or computational using problems But it’s not too early to demonstrate the benefits of solving of solving the benefits demonstrate to too early not it’s But discuss the merits of computer science as a career, he says. says. he sciencecareer, a as computer of discuss the merits FEATURE

she says. “It’s almost easier to disregard people who are know, but I feel so connected to it because of Women@ doing something malicious because they’re trying to get SCS,” she says. “It’s made me feel like I’m not an outsider.” to you, than to work with someone who you like, and try Marcu says that through the activities promoted by to tell them that something they intrinsically believe is Women@SCS, the group practices the inclusiveness it messed up.” preaches. “Men are involved in Women@SCS—highly In computer science culture, women who wear makeup involved—in things we do,” she says. “I’m not interested in or seem “pretty” can be perceived as being less serious having ‘more women’ in computer science, I’m interested about their work—including among other women. “Even in diversifying it overall.” when women feel comfortable being feminine, the extra attention they get for looking different than the norm is still a reminder of that feeling they don’t belong,” Marcu says. What’s the payoff? “It’s nice when people compliment me on my appearance, Does diversity in both gender and race lead to smarter but it makes me feel like an alien sometimes. It’s like you technologies and computing products? It’s hard to say, put on a decent set of heels and a skirt and everyone’s like, although there’s anecdotal evidence that the myopia caused ‘Holy cow.’” by lack of diversity can lead to product flops. Some of the e-commerce shopping sites that went bust during the early 2000s dot-com boom were designed from the perspective Building on the positives of male engineers, Ensmenger says. “They were focused As important as it may be to identify and root out gender on creating the least difficult means possible of acquiring or racial disparities, students say it’s also important not something, and that’s not what shopping is for most people. to spend too much time looking for negatives. Marcu It can be fun, it can be social. People wanted more out of attended one conference for women in computer science (e-commerce) than just what a male engineer might want.” in which participants told horror stories about sexual There are more recent examples: In 2009, reports circulated harassment, pay inequities and mistreatment. “It was just that cameras with face-detection technology were having lots of negativity,” Marcu says. “It felt like they were trying a hard time detecting the faces of dark-skinned people. to scare women out of going into computer science.” Last year, venture capitalist Marc Andreessen, co-creator Women@SCS “is not a support group for people to of the Mosaic browser and co-founder of Netscape, told complain about sexism in computer science,” Quispe says, Fortune Magazine that he ignored the potential of Pinterest “but it is definitely an outlet for me to feel better about until a female researcher urged him to spend some time being a woman in computer science.” with the site. His firm then invested $27 million. Too many computer technologies, Andreessen noted ruefully, have Word about Women@SCS has spread, says Frieze, who been “initially aimed primarily at men.” has shared material developed at CMU with universities throughout the United States and around the world. “CMU “If you’re going to have successful social networking has a reputation for paying attention to gender balance, but technologies, e-commerce, medical diagnostics, then we one of the things I have to always convince them of is that need to have systems designed for a broader perspective,” we don’t do anything special for women,” she says. That Ensmenger says. “If we have systems designed not only by philosophy drew Marcu into the group in the first place. males, but by a particular kind of white male who defines “I believe so strongly in what Carol is doing,” says Marcu, himself in terms of his ‘nerdiness,’ we’re going to have who served as coordinator of SCS’s Graduate Women’s problems.” Mentoring Program and currently speaks at Women@ Computer scientists often talk of the potential of their SCS roadshows. “Networking benefits women, not so that work to change the world. If that’s true, then, logically we can get together and talk about ‘women’s issues,’ but speaking, changing the population of computer scientists because seeing people who look like you is very important.” will eventually have a multiplier effect that will ripple Hearing prominent female computer scientists talk about through society. Women “need not only to be a part of their experiences is important as well, she says: “It helps me the culture, but contribute to shaping the culture,” Frieze envision someday being in their place.” says. “The idea is not simply diversity for diversity’s sake or For Schervish, Women@SCS is not just her link to female diversity for women’s sake. It’s become clear that diversity is faculty and students—it’s her link to her college and to the important to the industry.” university. “There are so many people in SCS that I don’t —Jason Togyer (DC’96) is editor of The Link.

24 Research notebook Polly spreads the word: Delivering information to a low-literacy audience using viral methods

By Agha Ali Raza and Roni Rosenfeld, Language This experience demonstrated that it is possible to virally Technologies Institute, CMU and Farhan Ul Haq, spread a speech-based service in a population with low Zain Tariq, Mansoor Pervaiz, Samia Razaq and Umar Saif, Lahore University of Management Sciences, literacy skills, using entertainment as a motivation. We Lahore, Pakistan then asked: can we leverage the power of entertainment to reach a large number of illiterate people with a speech- Editor’s Note: Portions of this article are adapted from the based service that offered a “payload”—a valuable service paper “Job Opportunities Through Entertainment: Virally such as: Spread Speech-Based Services for Low-Literate Users,” to • Facilitating an efficient marketplace (a speech-based be presented at the ACM SIGCHI Conference on Human “Craigslist”); Factors in Computing Systems, where it will receive a Best • Facilitating social and political activism (speech-based Paper Award. message boards and blogs); The ACM SIGCHI Conference on Human Factors in • Sending and receiving group messages (speech-based Computing Systems will be held April 27 to May 2 in Paris, mailing lists); and France (http://chi2013.acm.org). • Citizen journalism The original paper is copyright © 2013 ACM, all rights reserved, and may be downloaded at All of these services are available in written form via the www.cs.cmu.edu/~Polly/. web or text-messaging services. But very few such services are currently available to people with low literacy skills. The Link gratefully acknowledges the help of the ACM Copyright and Permission Team as well as the researchers working on Polly. In 2011, we developed a simple-to-use, telephone-based entertainment service called Polly that allowed any caller to record a short message, choose from several entertaining voice manipulations, and forward the manipulated recording to their friends. Introduced among low- skilled office workers in Lahore, Pakistan, within three weeks Polly had spread to 2,000 users and logged 10,000 calls. We eventually shut it down due to insufficient telephone capacity and unsustainable cost. In analyzing the traffic, we found that Polly was used not only for entertainment but also as voicemail and for group messaging, and that Polly’s viral spread crossed gender the and age boundaries—but not socio-economic ones. . link

25 Research notebook

Previous attempts The viral loop, and the power Several attempts have been made to design user interfaces of entertainment for users with limited reading and writing skills who are Ed Baker, the co-founder and CEO of Friend.ly, has inexperienced with technology. described what he calls the “viral loop.” To spread virally, For instance, in 2007, some of the same researchers on the according to Baker, a technology must create a catchy call to Polly team created Project HealthLine. Their goal was to action; create incentives for users to invite their friends; and provide telephone-based access to reliable spoken health maximize the average number of invites that users make to information for low-literate community health workers in a unlock those incentives. rural province of Pakistan. The speech interface performed A successful example of a cellphone-based service that has well but required training via human-guided tutorials. spread virally is SMS-all, a group text-messaging service Avaaj Otalo (2010) is another successful example of a in Pakistan. People use SMS-all to share information and speech interface serving users with limited literacy—in discuss hobbies and other interests. As of last report, the this case, farmers. Avaaj Otalo is a telephone-based system service had more than 2 million users and 400,000 groups, that offered an open forum where users could post and and had delivered more than 3 billion messages. answer questions; a top-down announcement board; In 2010, Thomas N. Smyth, Satish Kumar, Indrani Medhi and an archive of previously broadcast radio shows. The and Kentaro Toyama provided a demonstration of the most popular service turned out to be the open forum, power of entertainment. In their paper “Where There’s a constituting 60 percent of the total traffic. Users found other Will, There’s a Way: Mobile Media Sharing in Urban India,” interesting but unintended uses for Avaaj Otala, such as they describe a thriving black market in entertainment business consulting and advertisements. media on low-cost mobile phones, and highlight the But such projects have typically required explicit user remarkable ingenuity exhibited by users when they are training. As a result, they’re restricted to a moderate motivated by the desire to be entertained. number of users. When dealing with a large user base, Armed with these examples, we set out to develop an explicit training is not feasible. One alternative to such entertainment-driven telephone-based service that would training is to rely on learning from peers, and on spread virally among users with low-literacy skills. viral spread. Voice-based media can also promote social inclusion Introducing Polly among underserved communities. In 2012, Preeti Mudliar, Jonathan Donner and William Thies examined Polly is a telephone-based, voice-based application that participation in citizen journalism among rural allows users to make a short recording of their voice, communities in India using CGNet Swara, an interactive modify it and send the modified version to friends. In voice forum. CGNet Swara enabled users to record and Urdu, Polly is called “Miyan Mithu,” which has a meaning listen to messages of local interest and became popular similar to “Polly the Parrot.” among the target audience. With CGNet Swara, only Manipulations offered by Polly include converting a voice an initial set of a few dozen users received training. The from “male to female” or “female to male” by altering the rest of the users learned about it by word of mouth. pitch; a “chipmunk” effect, achieved with pitch and tempo modifications; an “I-have-to-run-to-the-bathroom” effect, achieved by a gradual pitch increase; converting the voice to a “whisper”; or adding background music.

26 the link. 27

he experiment he calls to five of the most frequent callers from Polly’s 2011 Polly’s callers from most of frequent the calls five to was back Polly that announced Thesebriefly calls study. was the system further to promote attempt No online. then since been continuously has up Polly ever made. minimal interruptions. with 9 and between May operations Polly of 130 days studied We than more had period,Sept. 15, 2012. During that Polly Most 85,000 users. with interactions 495,000 telephone days. few a times a few over only Polly with interacted users 160 seconds. was callduration Average job 530 suitable of a total Duringthis period, recorded we listened were per These 28 ads ads week. of average an ads, more to equates That user initiative. times—all 279,000 to by the number than more per possibly ad, 525 playbacks than were ads Some ad in a newspaper. that who read people of ad was popular most others—the than more much heard listened were 73 ads 8,400 times, and than more to listened 1,000 times each. than more to a particular deliver to made were 23,288 requests of A total to delivered were job ads Such user. another job ad to half than more interactions, such users. Of 9,475 different send they is, job ad delivery—that didn’t only requested the user probably that message—indicating Polly a “fun” ad the with job service. specificallyto interact called Polly and per $0.023 minute, of airtime cost outgoing an With Polly minutes, three about lasting interactions average with We peakairtime in alone. $400 per its us cost day during because option the toll-free eliminate to want didn’t T automated placing 9, 2012, by seeded was May on Polly olly works olly P

Polly is a telephone-based, voice-based application that allows users to to users allows that application voice-based telephone-based, a is Polly send modify the modifiedand it their voice, of recording short a make a which has Mithu,” called is “Miyan Polly Urdu, friends.to In version the Parrot.” “Polly to similar meaning

phone call itself. phone the system. Polly’s phone number is also played during the during also is played number phone Polly’s the system. all of Polly’s recipients on their first two interactions with interactions two their first on recipients Polly’s all of messages containing Polly’s contact information are sent to to sent are information contact Polly’s containing messages job ads. As an additional mechanism for viral spread, text text viralspread, for mechanism additional an As job ads. recording to others, create their own recordings or listen to to listen or recordings their own create others, to recording The recipient of a message can record a reply, forward forward the reply, a record can message of a recipient The browsed and forwarded to friends. to forwarded and browsed to the latest job ads. Those ads can be skipped, repeated, repeated, skipped, Those job ads. can be ads the latest to the recording to friends, give feedback about Polly or listen listen or Polly about feedback friends, to give the recording the recording again, re-record, tryforward effect, another re-record, again, the recording played back. The user is then offered the options to to hear options the offered user is The back. then played funny voice transformation of the recording is immediately immediately is therecording of transformation voice funny Pakistani newspapers and recorded them in Urdu. recorded and newspapers Pakistani A their voices. of recordings short make to prompted literacy workers. We collected such job descriptions from from job descriptions collected such We workers. literacy are users in the interaction, Early number. caller-paid descriptions of job openings suitable for low-skilled, low- low-skilled, for suitable openings job of descriptions a or be can via number made a toll-free Calls Polly to we call “the payload.” This was the option to browse audio browse to option was This the payload.” callwe “the How development-related application to Polly’s menu—what menu—what Polly’s to application development-related The most important advance was the addition of a addition was the advance important most The also were added. also were Extensive call logging and real-time monitoring capacity capacity call monitoring real-time logging and Extensive In addition, we allowed users to skip many of the menus. the menus. of many skip to users allowed we addition, In call airtime from $0.126 per minute to $0.023 per minute. $0.023 per to minute. $0.126 percallminute airtime from on one server in Pakistan, reducing the cost of our outgoing outgoing our of the cost server reducing one on in Pakistan, to 30 concurrent phone calls; and concentrated the software the software concentrated calls; and phone concurrent 30 to changes, we increased our telephone capacity to handle up up handle to capacity telephone our increased we changes, upgraded from the one first deployed in 2011. Among other Among in 2011. deployed first the one from upgraded The system we used in the current study was substantially substantially was study usedwe in the current system The Research notebook

that would bias the user base away from the low-income Conclusion people who are our prime target. So we experimented This was our first attempt to add a development-focused with limiting the number of times users could call the service—a “payload”—to Polly’s offering. We found that toll-free service per day. Each time we tightened the users took to the new offering in large numbers, and that quota, usage of the caller-paid service spiked. many started calling Polly specifically for job information— exactly the result we hoped for. Surveying users Why did only a handful of the users we interviewed After the study period, we called 207 randomly chosen apply for jobs? We attribute this to lack of trust or lack of users of Polly to collect demographic information. Out interest. We believe we can correct the “lack of trust” by of these calls, 106 resulted in useful information of some co-branding the ads with the names of familiar government type. Although 57 percent of respondents browsed Polly’s organizations or newspapers. We believe we can correct the job ads, only a handful reported applying for those jobs. “lack of interest” by exploring more job types (e.g., jobs for Two users claimed that their friends got jobs through the handicapped) to serve a wider selection of people. Polly, but we were unable to verify this. Usage of Polly grew exponentially because every user, on Respondents who described their primary use of Polly average, spread the system to more than one new user. Since as “fun” gave examples such as making prank calls to the target population is measured in the tens or hundreds friends, sending birthday greetings or hello messages, and of millions, volume will grow exponentially for quite a long browsing job ads as a pastime. Four sight-impaired users time, limited only by the system’s carrying capacity. defined Polly as an “alternative to text messaging” and However, usage over time is marked by rapidly declining praised it profusely. interest among most users. This was expected given the Among positive feedback for the job ad service, one man unchanging nature of the entertainment. In the short term, said (loosely translated) “after all that is going wrong with we are working to increase repeat usage by varying the the country…well, at least we have Polly…God bless Polly entertainment content. In the long term, we believe that a and may the service continue forever.” diverse “payload” of useful services will retain users, with entertainment being used mainly to spread to new ones. Is the service cost-effective? The jury is still out. Cost efficiency can be achieved not only by getting the users to pay for airtime, but also by the use of ads, carrier revenue- sharing and/or content sponsors (e.g., governments or NGOs). We are exploring all of these options.

Usage of Polly grew exponentially because every user, on average, spread the system to more than one new user. Since the target population is measured in the tens or hundreds of millions, volume will grow exponentially for quite a long time, limited only by the system’s carrying capacity.

28 the link. 29

Larry ECE andy andy Mutlu Onur Dean R erek alumni Derek Bryant, and (CS’93) Beatty and Huang (E’91), professor Left: SCS

2000s, but often we get people who graduated in 1970sthe get who graduated we people often 2000s, but degree. a “math-CS” ’80s with or a I met Texas, in Austin, brunch 24 alumni March our At Antonio San from 1967 CIT electrical alumnus engineering then called was what thewho been “digital had of member a group). circuits” the “analog to opposed (as group circuits” old years only two was Science Department The Computer our As degrees. graduate granted only it and point, that at more was group circuits” the “digital it, recounting was alum part himself of considers he therefore and “computer-y,” “analog and “digital circuits” distinction between SCS. (The us.) told he year, the following eliminated was circuits”

Science Computer chool of School he T turns 25 this year. Although SCS Although 25 this year. turns seven CMU’s of the youngest is back history stretches its colleges, intertwined is the 1950s, and to schools, and colleges other our with Tepper. and MCS CIT, including computer for events our a result, As interesting an are science alumni ustin, A and SCS Carr M. alumni arch 24 in 24 March Tina

exas. withyou

By Building our legacy our Building undergraduate or graduate alumni from the 1990s and the1990s and from alumni graduate or undergraduate mix of people—many of the alumni who attend are are who attend the alumni of people—many mix of on attended our attended alumni luncheon ECE T bout 60 people Right: About 60 alumni

It’s really rewarding to hear those kinds of stories. When our Generally speaking, when our alumni attend events, they alumni get together, we hear a lot of different professional have a lot of questions—have there been changes in classes experiences as well as different life experiences. I’ve now or courses? Do they still have 15-212 and 15-213? What been privileged to serve as the SCS alumni director for 13 new faculty have we added? Who’s still there? years, and I’ve been fortunate to watch our alumni grow But other times, they really just want to enjoy some “face both professionally and personally. It’s amazing to meet time” with one another and say, “Hey, how have you been, someone as a new undergraduate, see them wrap up a I haven’t seen you since then.” These days, they’re likely Ph.D., and then return to an alumni event with their spouse to be connected through professional organizations or and children. through social media networks such as LinkedIn, Twitter, By the way: Most of our events are family-friendly. We’ve Facebook, Tumblr, etc., but they’re always looking to had events at science centers and art museums, as well as grow their connections. So many of them are working on boat cruises and barbecues by the ocean. startups or hiring people for their businesses that alumni connections are a fast way for them to reach out and find Usually, both SCS dean Randy Bryant and ECE department qualified people. head Ed Schlesinger attend our events, but we also have special faculty speakers. In Austin, we hosted Onur Mutlu, Incidentally, that’s one of the reasons we invite and CMU’s Dr. William D. and Nancy W. Strecker Early Career encourage current CMU students to attend our summer Professor of Electrical and Computer Engineering. Onur events. If they haven’t graduated yet, students don’t always gave a very accessible talk about his current research in understand why alumni associations are important. We like computer architecture and industry partnerships that was to get current students involved so that they can meet our interesting and well received. graduates and say, “Hey, this is a really diverse group, a cool bunch of people, who do a lot of different things.”

Alumni By Geography Domestic*—T­ op Ten (by city) international*—Top eight (by country)

Princeton, NJ—85 Chicago—80 Taiwan R.O.C.—38 Philadelphia—98 Japan—41 Los Angeles—186 Canada—45 Washington, D.C.—236 Portugal— San Francisco 55 Bay Area—1,311 Boston—316 India—267 Singapore— 59 Seattle— 376 People’s Republic of New York Pittsburgh— China—98 South Korea— City—468 695 183

*As of October 2012 *As of October 2012 30 the link. 31

By now, you should have received two received have should you now, By your feedback makes a big difference when difference a big makes feedback your we’re planning new events and services for our alumni. services alumni. and our new events for planning we’re Please a comment? or idea, a suggestion an have Do you your hearing to I lookforward And a line. me drop town! in your time we’re the next stories M. Carr (HNZ’02) Tina Director of Alumni Relations Alumni of Director [email protected] installments of our new alumni e-newsletter, “Bytesize.” “Bytesize.” e-newsletter, new alumni our of installments online in CMU’s register to sure make haven’t, you If current send your least me at or community, alumni sowe’re email, of a lot get you know We address. email kept we’ve why that’s you—and trying spam to not bite-size. well, “Bytesize”… LinkedIn this, our read the time you else? By What this running. With and beup should community the SCS with connect to be able you’ll new addition, media platforms—Facebook, major four on community us leave should Google+ That LinkedIn. and Twitter, invents alumni our of one (until covered well pretty socialanother media technology). s always, As , so what’s new? what’s so OK,

*As of October 2012 Male: 5,153 Male: At-A-Glance* Female: 1,193 Female: we’re also connecting alumni with prospective prospective with also alumni connecting we’re

Alumni hese days, days, hese SCS 6,346 Alumni: Total T maintaining and expanding our tradition of excellence. of tradition our expanding and maintaining to attend SCS. Fostering these legacies will be important to to these legacies will be important SCS. Fostering attend to students who are now the second generation of their family their family of generation the second now who are students quarter-century, we’re starting to develop “legacies”— develop to starting we’re quarter-century, As the School of Computer Science enters its second its Science enters Computer the School of As with someone.) this is a nice way to do it. Email me and we’ll connect you you connect we’ll and me Email do it. to this a nice way is for an alumni activity that’s fun and mutually beneficial, mutually fun and that’s activity alumni an for path. (Incidentally, if you’re looking for a way to volunteer volunteer to a way for looking if you’re (Incidentally, path. did you work with, and how did you choose your career career choose your did you how and with, work did you to ask questions such as what was CMU like, what faculty what CMU like, was what as such questions ask to connecting them to recent Ph.D. alums, encouraging them encouraging alums, Ph.D. recent them to connecting We do the same thing for prospective graduate students— graduate prospective for thing do the same We if it’s appropriate for them.) for appropriate if it’s attend SCS. (Of course, we want them to choose themto CMU— want we SCS. (Of course, attend experience is like, and think about what it would be to like would it what think about and experience like, is them to talk to recent graduates about what the student the student what about graduates talkthem to recent to to CMU but haven’t yet decided to attend. We encourage encourage We decided attend. yet to haven’t CMU but to Bay area for high school students who have been accepted who have highstudents school for area Bay past few years, we’ve held receptions in the San Francisco Francisco in theSan receptions held we’ve years, few past students who are thinking about attending CMU. For the For CMU. attending thinkingabout who are students on campus giving back: Pradeep Sindhu

By Mark Dorgan

As founder, vice chairman and chief technology officer of Juniper Networks, Pradeep Sindhu has his eyes on sustainable business models and the bottom line. But Sindhu (CS’83, ’84) remains a computer scientist at heart, with an interest both in the design of complex systems and in the work of the School of Computer Science. Recipient of an alumni achievement award in 2008, Sindhu says he values the connections he maintains with SCS. In his role as Juniper’s CTO, Sindhu is responsible for outlining the company’s technical needs and developing future projects. Among the most important benefits of his CMU education was the foundation it provided for his career—a solid grounding both in the fundamentals of computer science and the technologies behind computer systems. His involvement runs deep at SCS, from supporting a graduate fellowship, to hosting CMU’s Silicon Valley Network Night events at Juniper’s Santa Clara facilities in 2012 and 2013, to speaking on campus and facilitating delivery of Juniper’s networking equipment to the Gates involvement for two consecutive years is a direct result of and Hillman Centers. Sindhu’s interest. Sindhu says his involvement is motivated by a desire “CMU depends on its alumni for continued success,” to give back and share his experiences. That’s been Sindhu says. “I feel strongly that alumni owe it to manifested in gifts such as the one that created the CMU and SCS to stay involved and provide Pradeep Sindhu Graduate Fellowship to provide philanthropic support.” support to a graduate student in the Computer Science Nearly 48,000 donors have made almost 222,000 Department. “It’s a great way to stay in touch with students individual gifts to the university as a part of Carnegie who are experiencing the rewards and challenges of Mellon’s “Inspire Innovation” Campaign, which has now earning a Ph.D.,” he says. raised $1.11 billion. About $560.3 million has gone into Severin Hacker, a recipient of the Sindhu fellowship, the university’s endowment. had the opportunity to meet Sindhu during his visit to Until June 30, you can be a part of this historic milestone campus in the fall of 2011. “Pradeep Sindhu stands not for CMU. To find out how you can help the School of only for outstanding technological achievement, but Computer Science through scholarships, fellowships, also exceptional entrepreneurial achievement,” Hacker faculty support or gifts to the Gates and Hillman says. “Given Pradeep’s professional and technological Centers, please contact me at [email protected] or accomplishments, I’m very grateful to receive the call me at 412-268-8576. You can learn more about the Sindhu Fellowship.” Inspire Innovation campaign by visiting www.cmu.edu/ Hosting Network Nights provides Sindhu with a way to campaign. help CMU build its presence in Silicon Valley—where —Mark Dorgan is executive director of major gifts and a large number of alumni work and live—and promote development liaison for the School of Computer Science. the university’s Southern California campus. Juniper’s

32 the link. 33

a sport she enjoyed while at CMU, when she completed completed when she CMU, while at enjoyed a sport she in did a half-marathon Tomko marathons. four this later do a full to marathon hoping is and November in every state,” run to marathon a like I’d “Ideally, year. Pittsburgh: misses Tomko the reasons of one It’s says. she to.” easily drive I could that states more many had “I (DC’96) Togyer —Jason “it canceled the speech interaction, not the destination,” the destination,” not the speech canceled interaction, “it says. Tomko focus of a major scenarios is those of kinds Anticipating we goes wrong, something “When says: she team, her possible.” as minimally frustrating as it need make to was “I science via linguistics. computer to came Tomko answer,” ‘right’ any never was there and major, English an it argument, a coherent made you as long “As says. she I had sets, and problem had we linguistics, In OK. was things doing like ‘I I thought: and them out, work to the end.’” at a problem solve and hard work you where in Ph.D. and master’s earn her to CMU to came Tomko worked she where Institute, Technologies the Language language of CMU professor Rosenfeld, Roni with science, computer and learning machine technologies, user to respond could that speech interface a universal on input the user’s of the syntax then shape and interactions responses. own its varyingof theby syntax had always because he adviser, a great was “Roni for suggestions interesting and problems challenging also was “He says. Tomko things,” thinking about of ways opportunities, networking very providing good about good it was to the field, into getting just a student as and ‘Roni say, and out reach could you whom to people have us.’” introduced ECEand SCS many with works Tomko Microsoft, At The (E’01). Gaugel Dan husband, her including alumni, Alex their children with Seattle in suburban lives couple national in the moment had a Tomko Theo (5). (3) and $43,000 than more won 2009, when in she spotlight show game the syndicated on champion a two-day as running again— up took recently she And “Jeopardy!” Stefanie Tomko Tomko Stefanie speechdesigns systems. recognition bit a little So it’s when she surprising have don’t “You says, control speech to Do your everything.” you? deceive ears right. heard you No, senior Tomko, manager program in Microsoft’s Embedded Windows snapshots alumni

B.A., linguistics, University of University linguistics, B.A., 1996 Washington, 2001 Carnegie University, Mellon science, computer M.S., 2007 Carnegie University, Mellon technology, language Ph.D.,

tested. When a beta user said “cancel” to erase a destination, erase a destination, to a betauser When “cancel” said tested. cliff.” She describes a navigation system currently being beta currently system describesShe navigation a cliff.” I say, but that when it doesn’t, the user doesn’t fall off of a fall off theuser doesn’t doesn’t, when it that but I say, it intuitive—not necessarily that it understands everything understands it that necessarily intuitive—not it challenge,” Tomko says. “The other big challenge is makingis challenge “Thebig other says. Tomko challenge,” dealing with nonnative speakers of a language is a big a big is a language of speakers dealing nonnative with she says, but difficult problems remain. “Right now, now, “Right remain. problems difficult but says, she filtering ambient noise—have been “reasonably solved,” solved,” “reasonably been noise—have ambient filtering Many of the problems of speech recognition—such as speech recognition—such of the problems of Many can’t just be tacked on top.” be on tacked just can’t make sure it ‘plays nice’ with the visual user interface. It It the visual with userinterface. nice’ ‘plays it sure make has to be designed to minimize distractions. We need to We be to minimize distractions. designedhas to useful it is,” Tomko says. “In a vehicle, speech recognition speech recognition a vehicle, “In says. Tomko useful is,” it interaction is just right. “The more it’s ‘blabbering,’ less the ‘blabbering,’ it’s “The right. more just is interaction Tomko and her team. Their focus is making sure that the that focussure making is Their team. her and Tomko climate control, you’ve seen and heard the work done by by done the work heard seen and you’ve control, climate activated controls for phones, music, navigation or or navigation music, phones, for controls activated If you’ve ridden in a late-model Ford or Kia with voice- with Kia or Ford ridden in a late-model you’ve If going to increase it.” increase to going down,’ and you don’t know exactly how much the system is the system much exactly how know don’t you and down,’ now you have to say something like, ‘volume up’ or ‘volume ‘volume or up’ ‘volume like, something say to have you now volume control is a terrible thing to use speech for, because use speech to thing for, a terrible is control volume turn it one way to go up and the other way to go down. But But down. go to way the other and up go to way one it turn around forever, and everyone knows how it works. You You works. it how everyone knows and forever, around “Take your radio volume control,” she says. “It’s been “It’s says. she control,” radio volume your “Take advantages over traditional knob-and-button interfaces. interfaces. knob-and-button traditional over advantages recognition, or voice-command, systems need to offer clear need offer to systems voice-command, or recognition, Connected Car Technology Group, says speech says Group, Connected Car Technology Stefanie Tomko Stefanie alumni snapshots Séan Slattery B.A., computer science, Trinity College, Dublin, 1993 M.S., computer science, Carnegie Mellon University, 1998 Ph.D., computer science, Carnegie Mellon University, 2001

known as derivatives, and then minimizing the exposure to risk for the company and its clients. “We’re writing the core calculations that get slotted into other pieces of software,” he says. “We’re running these calculation engines by writing formulae that require a really deep understanding of how financial maths work. There’s a lot of software in investment banking these days, and it requires a lot of attention to detail—to ‘getting it right.’ It also requires very strong programmers.” The basic skills that Slattery learned while working on his doctorate in computer science are the skills he uses every day, he says, adding that Tom Mitchell, CMU’s Fredkin Professor of Artificial Intelligence and head of the Machine Learning Department, was an important influence. “One of the things I really admired about Tom is that he was A career in the financial sector? For a guy who grew up very, very good at cutting to the core of any problem and in the small city of Port Laoise in the Irish Midlands (“it defining what the important pieces were, and what the was kind of like the U.S. Midwest—think Ohio”) the idea unimportant distractions were,” Slattery says. left Séan Slattery cold. “It’s got this kind of Hollywood Slattery earned his undergraduate degree at Dublin’s reputation, and I had visions of people on these high- Trinity College and was attracted to CMU because one of pressure trading floors, shouting at one another,” he says. his classmates, Joseph O’Sullivan (CS’97), had gone there. Instead, when Slattery finished his Ph.D. at CMU, he went “I knew it was a tough school, but we didn’t quite have the to work at a startup in London that was developing a new Web then, so I had written a letter asking for literature on search engine. “I hung on there for about two years, but the master’s program,” he says. “They wrote back and said, when they started having redundancies, I decided it was well, we don’t have a stand-alone master’s program right time for me to start looking around,” he says. He ended up now, but why don’t you apply for the Ph.D. program?” in “The City”—London’s financial sector—where he found He was glad he did. “Both Carnegie Mellon and Pittsburgh a different environment than he’d envisioned. It appealed were good places to be,” Slattery says. “Pittsburgh is a to his love of problem solving, and puzzles. wonderful place. It’s a proper city, but it’s a very manageable “When I was a kid, I used to go to the library, and when scale, and then Carnegie Mellon had a family atmosphere I was finally allowed to join the adult library, I discovered that was very encouraging to my research.” all of these books of puzzles by Martin Gardner from Slattery’s busy schedule doesn’t leave a lot of room for Scientific American,” he says. “I loved them. I ended up hobbies, though he and his wife, Gordana, do have a tearing through them, and that’s when I first realized I passion for cinema. “My tastes range from classics like had an aptitude for that kind of thinking.” ‘The Third Man’ to the kinds of stuff that you would In his current role as head of emerging markets, classify as ‘bad movies,’” he says. commodities and FX quants, Americas, for Credit Suisse —Jason Togyer (DC’96) Group AG, the puzzles that Slattery’s group is solving require deriving the fair prices for financial instruments

34 scs news in brief

Two with CMU ties win Turing Award

Two researchers with ties to both Carnegie Mellon University and SCS professor Manuel Blum have received the A.M. Turing Award from the Association for Computing Machinery.

Shafi Goldwasser (CS’79) and Silvio Micali, both affiliated with MIT’s Computer Science and Artificial Intelligence Laboratory, are co-recip- ients of the 2012 Turing Award for their work to take data encryption from theory to practice.

They will receive the award—some- times called the “Nobel Prize of computing”—at the ACM’s annual banquet June 15 in San Francisco. Goldwasser and Micali The Turing Award carries a $250,000 prize, with financial support from Berkeley, came to CMU in 1999 and European Association for Theoretical Intel and Google. currently serves as the university’s Computer Science and the ACM’s Goldwasser is a 1979 graduate of Bruce Nelson Professor of Computer Special Interest Group on Algorithms Carnegie Mellon with a B.S. in mathe- Science. and Computation Theory. Her 1993 Gödel was shared with a team that matics who earned her M.S. and In honoring Micali and Goldwasser, included Micali. Ph.D. in computer science from the the Turing Award committee said the University of California at Berkeley. pair “laid the foundations of modern Micali has been a regular visitor to With the win, Goldwasser becomes the theoretical cryptography, taking it the CMU campus. In February 2000, 12th CMU affiliate—including faculty from a field of heuristics and hopes to Micali presented a paper on “certified and alumni—to be honored with a a mathematical science with careful email” as part of the SCS Distin- Turing Award. Named for artificial definitions and security models.” guished Lecture Series, and in 2004 and 2011, Micali spoke at theory intelligence pioneer Alan Turing, the The results they have achieved, seminars held by CMU’s Center for very first Turing Award was presented working together and with others, “es- Algorithm Adaptation, Dissemination in 1966 to Alan J. Perlis, founding tablished the now-standard definitions and Integration. He currently serves head of the Computer Science Depart- of security” for encryption and digital as Ford Professor of Electrical Engi- ment at what was then the Carnegie signatures, the award committee said, neering and Computer Science at MIT. Institute of Technology. adding that Goldwasser and Micali Micali is a graduate of the University established the “tone and character Goldwasser is the RSA Professor of of Rome who also earned a Ph.D. of modern cryptographic research.” Electrical Engineering and Computer at Berkeley. Their work in collaboration with other Science at MIT and also a professor of researchers has “provided stunning computer science and applied math- Goldwasser and Micali were advised innovations,” the committee said. ematics at the Weizmann Institute of at Berkeley by Blum, himself a Turing Science in Israel. Award winner in 1995. Blum, formerly Goldwasser is a two-time winner of a professor of computer science at the Gödel Prize presented by the

SCS, Tepper centers join forces to help entrepreneurs

Two change agents on the CMU cam- Olympus was founded in January preneurship,” says Blum, who will pus have joined forces to help students 2007 by Lenore Blum, professor of co-direct CIE along with Tepper’s and faculty bring their ideas to market. computer science, and operates as Dave Mawhinney. “The knowledge part of SCS. The Don Jones Center and skills necessary to start a Project Olympus and the Donald H. was created by CMU’s Tepper School business don’t come naturally, Jones Center for Entrepreneurship are of Business. regardless of how gifted a person

partnering to form a new Carnegie the might be in their chosen discipline. Mellon Center for Innovation and “We want Carnegie Mellon to be the

The Center for Innovation . link Entrepreneurship. destination of choice for students and Continued on page 38 faculty who are interested in entre-

35 scs news in brief

Crawling CHIMP takes on next DARPA challenge

It looks like a primate, but crawls on focused on simplicity and reliability, tracks like a tank—the newest robot to which favors tracked locomotion. take on a DARPA challenge is the CMU Walking and standing greatly increases Highly Intelligent Mobile Platform, or the complexity of the processing CHIMP for short. power required by a robot, which in The Defense Advanced Research Proj- turn increases its power consumption, ect Agency Robotics Challenge seeks Stentz says. It also leaves the robot robots that can respond to search-and- more vulnerable to energy or program- rescue situations such as the 2011 ming glitches, he says. Fukushima nuclear plant disaster. CHIMP will be able to navigate through The robots will need to be able to hazardous situations and maintain navigate stairs and ladders as well stability automatically, while a human as operate equipment and vehicles operator working via remote control designed for humans. will make higher-level decisions. “This CHIMP is being developed by CMU’s enables CHIMP to be highly capable National Robotics Engineering Center, without the complexity associated whose entry, Tartan Rescue Team, is with a fully autonomous robot,” one of seven selected by DARPA for Stentz says. “track A” of the challenge. In addition to the Tartan Rescue Team, With four articulated limbs that a second CMU team is also competing incorporate hand-like claws, CHIMP in this DARPA challenge. Team Steel is does look something like a monkey headed by Chris Atkeson, professor of or chimpanzee. But NREC scientists robotics and human-computer interac- say its normal form of locomotion tion. Unlike Tartan Rescue, which is de- Events are planned won’t be walking on its legs, but rolling veloping both hardware and software, for this June and December, along tank-like tracks built into those Team Steel is developing software in a with the final event scheduled four limbs. virtual competition that will run on a for December 2014. The winner will DARPA-provided humanoid Atlas robot. receive $2 million. “This type of robot has tremendous potential,” says Tony Stentz, director Robots in the challenge will have to In 2007, Tartan Racing, a joint team of NREC and leader of Tartan Rescue successfully perform tasks such as fielded by Carnegie Mellon and General Team. Although the human-size tasks driving a vehicle, traveling across rub- Motors, won the $2 million DARPA required in this challenge would seem ble, removing debris, opening doors, Urban Challenge to develop a driver- to favor a robot that can stand and climbing ladders, and using tools to less vehicle that could successfully walk like a person, he says the team is operate and replace components. navigate an urban environment.

NREC’s robot paint-stripper an Edison Award finalist

A robotic system that would remove paint and other coatings from aircraft has been named a finalist for the internationally renowned Edison Awards.

The Advanced Robotic Laser Coating Removal System is being developed on behalf of the U.S. Air Force by CMU’s National Robotics Engineering Center in conjunction with Concurrent Technologies Inc. It uses high-powered lasers mounted on mobile robotic platforms.

Named for inventor Thomas Edison, the Edison Awards are presented annually in 12 different categories and are voted on by more than 3,000 senior business executives and scientists from across the United States. First presented in 1987, the awards honor innovation and creativity in product and service design.

36 the link. 37

Anderson May at the APS’s annual convention convention annual APS’s the at May Washington. in Anderson will receive the award in award the receive will Anderson Rebecca Balebako of Balebako Rebecca Departmentthe Alex ofPolicy, Public and Engineering ofBeutel De- Science Computer the partment,ofLing Wang Languagethe Meza Justin and Institute Technologies ofDepartmentthe of andElectrical Engineering. Computer Justin Cranshaw Justin John R. John receives

receive researcher honor The program was that successful so The CarnegieLearning was formed bringto half-milliona than More market. to it now have schools 2,600 at students Carnegie by developed software used Learning. oflot a been have “There well-inten- effortsunsuccessful but tioned at technology computer applying edu- to Bryant, dean,SCS Randy says cation,” Bydeal. real “the are Anderson’s but ofmodels developing students how technologytutoring learn,cognitive his successful.” remarkably been has waythe pointed research Anderson’s by done being that as such work to ofScience Pittsburgh the Learning ofproject joint a Center, theand CMU of University usingis which Pittsburgh, new develop both to software tutoring ofmethods datacollect and teaching classrooms.actual in students from Research, is working on research that research on working is Research, harnessescomput- mobile and social of lives the improve to applications ing cities. populated densely in people - graduate12 fellow awarded Facebook year. academic 2013–14 the for ships includedCMU from finalists Other Julian Shun Julian

tutoring students fellowships

highest

grad

acebook teach algebra to high school students. students. school high to algebra teach created a computer tutor that couldthat tutor computer a created In the 1990s, a team led by Anderson by led team a 1990s, the In used as instructional aids. instructional as used how computer-based systems can becan systems computer-based how the brain works, how people learnpeople andhow works, brain the computer science to understand how how understand to science computer from both cognitive psychology cognitive both from and faculty since 1978, combines research research combines 1978, since faculty Anderson, a member of member a Anderson, CMUthe Award for Basic Research. Basic for Award William James Lifetime Achievement Achievement Lifetime James William named this year’s winner ofwinner year’s this named APSthe and Computer Science, has beenhas Science, Computer and University Professor of Professor University Psychology John R. Anderson, CMU’s R.K. MellonR.K. CMU’s Anderson, R. John Psychological Science. Psychological highest honor of honor highest for Association the cognitive tutors has been awarded the awarded been has tutors cognitive creation of creation first Carnegie Learning’s The researcher whose work led toled work whose researcher The ’s society Cognitive student in the Institute for Software Software for Institute the in student Cranshaw, a fourth-yeara doctoral Cranshaw, programming. niques that simplify shared-memorysimplify that niques - tech as well as processing parallel is working on algorithms for large-scale large-scale for is working on algorithms connected through technology, a technology, through connected Shun computers. multi-core memory, whose research focuses on shared shared on focuses research whose Facebook Graduate Fellowships Fellowships Graduate Facebook DepartmentScience Computer the in will each receive a $30,000 stipend.$30,000 a receive each will andopen more world the make to fourth-yeara is Shun student doctoral Justin Cranshaw and Julian Shun alsoShun Julian and Cranshaw Justin waysinto supportresearch emerging said. spokesman company and other fees. other and their tuition, their travel to conferences conferences to travel their tuition, their fellowships by Facebook that pay for pay that Facebook by fellowships Two Ph.D. students have been awarded awarded been have students Ph.D. Two

Two F scs news in brief

CSD’s Miller wins 2013 Knuth prize Entrepreneurs Continued from page 35 the same except for the labeling of their components. In 1984, Miller and Entrepreneurship provides the moved into the area of scientific missing pieces, benefitting not only computing. He set up the theoretical our faculty and students, but the foundations for mesh generation, and entire region.” his subsequent research with Ioannis Koutis and Richard Peng into solving Since its creation, Olympus has “symmetric diagonally dominant” supported more than 100 projects linear systems has had important developed by CMU students and faculty. applications in image processing, More than 70 companies have been network algorithms, engineering and launched, attracting $60 million in physical simulations. additional funding from university and outside sources. Before coming to Carnegie Mellon, where he serves as professor of Four of those companies received computer science, Miller held venture capital funding, and one— faculty positions at the University BlackLocus—raised $35 million. Some of Waterloo, the University of 250 students have participated in Gary L. Miller Rochester, Massachusetts Institute product development supported in of Technology and the University of part by Olympus. Southern California. Gary L. Miller has been honored “The Don Jones Center and Project with the 2013 Donald E. Knuth Prize Miller was made a fellow of the Olympus have been tremendously suc- for his contributions to the fields of Association for Computing Machinery cessful in helping students and faculty cryptography, number theory, parallel in 2002 and in 2003 shared the ACM take the crucial first steps in transform- computing, graph theory and related Paris Kanellakis Award with three other ing research findings into products and fields in theoretical computer science. researchers for the Miller-Rabin test. services that people want to buy,” says Among his Ph.D. students are Susan Mark Kamlet, CMU provost and execu- Miller received his Ph.D. from the Landau, Tom Leighton, Shang-Hua tive vice president. Kamlet will serve as University of California at Berkeley in Teng and Jonathan Shewchuk. head of CIE’s new governing body. 1975 under the direction of Manuel Blum. In a paper published that same The Knuth Prize is presented jointly CIE will continue successful programs year, Miller introduced the first efficient by the ACM’s Special Interest Group from both the Don Jones Center and algorithm to test whether a number is on Algorithms and Computation Project Olympus, including Olympus’ prime. The generation of large prime seasonal Show & Tell events that con- Theory and the IEEE’s Computer screenshot numbers is an essential part of the Society Technical Committee on nect campus researchers with local and RSA public key cryptosystem, on the Mathematical Foundations of national investors; the Open Field En- which much of today’s Internet Computing. The prize is named in trepreneurs Fund; incubator space for commerce depends. honor of Donald Knuth of Stanford fledgling companies; entrepreneurship University, who has been called the workshops; and business competitions. In 1976, Michael O. Rabin showed father of the analysis of algorithms. how randomization could be used to New this year is a program called It is given annually by SIGACT and turn Miller’s algorithm into one whose “Launch CMU,” a series of events that TCMF, and includes a $5,000 award. efficiency did not rely on any unproven will bring top-tier venture capitalists hypotheses. The resulting Miller-Rabin Miller will be presented with the award together with the university’s most test is now the main method used in when he delivers the annual Knuth promising researchers and entrepre- practice for RSA encryption keys. Prize Lecture during the Symposium neurs twice a year in both Pittsburgh on Theory of Computing in Palo Alto, and Silicon Valley. The first event is Miller also made significant Calif., to be held June 1–4. slated for May 21 in California’s contributions to the theory of Silicon Valley. isomorphism testing—the problem of telling whether two structures are

38 the link. 39 cookie

’s website: www.cmu.edu/herb-robot/ website: HERB’s

It takes a tough robot to disassemble a fragile cookie. fragilecookie. a disassemble tough robot to a takes It By now, you’ve probably seen the video, which has been viewed more than a quarter- a than more viewed been has which video, the seen probably you’ve now, By of makers Homethe The HERB, recruited cookies Oreo YouTube. on times of-a-million to ofQuality Center, Pittsburgh’s at Technology developed Butler Life Robot Exploring partas filling cream their from cookies the of separate campaign advertising national a screenshot Creme.” Vs. “Cookie called of layers many programmers, are including HERB’s There video. that behind science of professor associate Siddhartharobotics Pras scientist project and (CS’05) Srinivasa periodtwo-week a over day per hour an about spent ’12), ’09, CS’05, (E’05, Velagapudi necessaryskills motor fine the in robot the training pryto apart cookies. delicate the photo.) the in Velagapudi (That’s partone task—in the with confronted when vexation (staged) HERB’s at laughed all We the ofseparate to unsuccessfully tries HERB “out-take,” an as staged that’s video the withthem smacking by then and knife, butcher a with them at stabbing by first cookies arms wellas work HERB’s how showing too, side, serious a had video fryinga The pan. world. visual its perceive to HERB allow that cameras the as ofproject joint a is of QoLT University the and CMU side. serious a has also HERB Pitts- challenges— physical with people help to designed technologies develops which burgh The homes. their in independently live adults—to older and includingveterans wounded explore that algorithms manipulation the validate to researchers allowed challenge Oreo - environ ordinaryin kitchen home, at people assist could robots other or HERB how tasks. everyday with ments, at Learnmore Bgfpl0lcIo ifAnd didn’t www.youtube.com/watch?v=g video: the see you Smart then and now then and now: computer science’s female pioneers A rchives ennsylvania P ennsylvania niversity of U niversity hoto courtesy courtesy P hoto

Do you think women were always a minority in After the war, Lichterman moved with ENIAC to the computer science? Think again. Army’s Aberdeen Proving Ground in Maryland, where she taught programming, while Snyder and Jennings You may already know about computing pioneers helped design and write code for the first UNIVAC. such as Grace Hopper and Ada Lovelace, but in McNulty married John Mauchly, co-inventor of the early days, women were highly visible at every ENIAC and UNIVAC, and collaborated with him on level of computer programming—including program design. entry-level coders. Through the 1950s and 1960s, computer programming In fact, the word “computer” originally applied to was a popular career path for women, but a variety the people (mainly women) who computed, by of social, political and economic factors led the field hand, the answers to complicated mathematical to become increasingly dominated by men. Some of equations. During World War II, more than 80 those same factors have also discouraged men and of these women “computers” were hired at the under-represented minorities from pursuing careers in University of Pennsylvania to plot ballistic missile computer science and information technology. trajectories using differential equations. Since the 1990s, Carnegie Mellon University has When Penn began work on the world’s first made a concerted push to encourage young women all-electronic digital computing machine, six of and under-represented minorities to explore careers those women—Frances Bilas, Jean Jennings, Ruth in computer science. In February, CMU joined 19 Lichterman, Kathleen McNulty, Frances Snyder other universities and more than a dozen corporations and Marlyn Wescoff—were asked to become its to participate in the Pacesetters program recently programmers. The machine was the Electronic launched by the National Center for Women & Numerical Integrator and Computer, or ENIAC. Information Technology. And every day, students and Working in teams, the women programmed faculty are engaged in outreach and networking, both ENIAC by manually routing cables and setting to attract people to computer science—and keep them switches and dials. engaged in the field. In this February 1946 photo from Penn’s archives, Read more about Women@SCS on page 16. Jennings and Bilas are shown preparing ENIAC for its public unveiling. —Jason Togyer (DC’96) 40 All events to be held on the Carnegie Mellon University campus in calendar of events Pittsburgh, unless otherwise noted. Dates and locations subject to change without notice. Visit calendar.cs.cmu.edu for a complete and current listing of events.

April 30 May 19 Aug. 26 2013 Celebration of Education SCS Diploma Ceremony Fall semester begins 4:30 p.m., Rangos 1 & 2, 1:30 p.m., Carnegie Music Hall, University Center 4400 Forbes Ave. Sept. 2 Labor Day: No classes May 2 May 23–25

Senior Student Leadership Workshop honoring Ravi Sept. 26–28 Awards Reception, Kannan’s 60th birthday Ceilidh: Homecoming Qatar Campus Various locations, and Family Weekend 6 p.m., Assembly Area, Pittsburgh campus Various locations Education City June 14–15 Oct. 18 May 3–5 Carnegie Mellon Alumni Mid-semester break: Conference on Automated Volunteer Forum No classes Personal Genome Analysis Pittsburgh for Clinical Advisors: Challenges and Solutions Oct. 24 July 13 Gates and Hillman Centers TechBridgeWorld Interactive SCS/ECE Alumni Reception 5 p.m., Perlis Atrium, Seattle May 6 Newell-Simon Hall Qatar Graduation Ceremony 2013 July 27 Nov. 18–22 7 p.m., Qatar National SCS/ECE Alumni Reception Spring 2014 registration week Convention Centre San Francisco Nov. 27–29 May 18–19 Aug. 18–25 Thanksgiving holiday: Commencement Weekend First-Year Orientation No classes Various locations, May 19 Pittsburgh campus Dec. 6 University Commencement Fall semester: Ceremony Aug. 21 Last day of classes 11 a.m., Gesling Stadium Qatar Convocation Class of 2017 Dec. 9–17 6:30 p.m., Education City Final exams the the . link

41 Office of the Dean 5000 Forbes Avenue Pittsburgh, PA 15213

Follow us!

SCSatCMU

We crossed the $1B mark. We connected with more alumni than ever. We helped grow CMU’s global visibility.

Thanks to all of you — nearly 50,000 alumni, faculty, students, parents and staff — who supported the Inspire Innovation campaign, we are in the homestretch of an historic effort that’s already impacting everything from new research facilities to faculty and student support.

But we’re not done yet. Help us continue the celebration as we prepare to cross the campaign finish line in June.

Celebrate with us at we.cmu.edu to add your gift today.