PPreetd ais cc ta in le g Rese Ca or m cp hu t Fu in tg ures

Next-Generation Research and Breakthrough Innovation

Indicators from US Academic Research
This article is based on a talk and an internal report that the author presented at faculty workshops and conferences. We offer it here to provide a glimpse into some of the activities that a major software company carries out to prepare for future changes in computing. Unlike our usual articleswhich are typically popular versions of research articles or news reportsthis piece reflects industrial thinking that is broad, concerned with the sociology of science, free flowing in associations, and developed from somewhat of a corporate viewpoint. Accordingly, we suggest readers view it as an op ed pieceone thats thought-provoking in part because of the authors position and activities. Rubin H. Landau, News department editor

hen searching for breakthrough advantages, the key innovations are those that surpass the present state so significantly that they might well lead to the next generation of technical advances. Over a period of a year and a half, I interviewed more than 100 educators, researchers, and deans in many disciplines with the overall goal of encouraging innovative collaborations between Microsoft Research and academics as part of my responsibilities as a university-relations specialist. The survey reveals connections among a broad range of topics and various recurrent themes that might be of interest to scientists in various fields. The interest of society, science, and industry in breakthroughs originates from either a need to quickly solve urgent problems or a desire to gain advantage over competitors. In todays richer countries, for example, manufacturing is less than

1521-9615/09/$26.00 2009 IEEE Copublished by the IEEE CS and the AIP

Thomas C. McMail
Microsoft Research

20 percent of economic activity, so breakthroughs in the knowledge economy are increasingly important. As The Economist points out, although the US still leads the world in research spending and innovation, its advantage over smaller and developing rivals is being eroded by the powerful leveling effects of globalization and information technologies.1 Yet, as the authors also note, this is not a bad thing because even if innovation is the key to global competitiveness, its not necessarily a zero sum game because the well of human ingenuity is bottomless, innovation strategies that tap into hitherto neglected intellectual capital and connect it better with financial capital can help both rich and poor countries prosper. That is starting to happen in the developing world. 2 Breakthroughs seem to create many transformational changes. However, by definition, breakthroughs are changes that advance us far beyond the present state, so its generally impossible to foresee the next big thing by simply taking the next logical step. As Nassim Taleb points out in The Black Swan, the inability to predict outliers implies the inability to predict the course of history.3 Accordingly, to increase the likelihood
Computing in Science & Engineering

76 

$1,000 $800

$600 $400 $200 $0 1970

Math and computer science Life sciences Engineering Physical sciences Environmental sciences Social sciences Psychology Other

Millions (US dollars)

1975

1995 1980 1985 1990 Research trends by discipline

2000

2005

Figure 1. US National Science Foundation funding for various disciplines over time (in constant fiscal year 2007 dollars). Other indicates unclassified basic and applied research, but doesnt include development and research and development facilities. As the figure shows, computer science continues to grow. (Courtesy The American Association for the Advancement of Science.)

of breakthroughs, he suggests that we should approach research much as we might an investment portfolio: focus on the next logical step, but also allocate 15 percent of resources to high-risk, potentially high-gain investments. In looking for breakthrough innovations in computation, at Microsoft Research we consider tools and techniques together because we view them as equally important. In particular, new and powerful tools tend to define both how and what work will be done. Furthermore, major computation breakthroughs tend to be team efforts: computer science (CS) team members focus on the tools, but collaborate with specialists on the science. Jim Gray offered a computer scientists inside look at this collaboration process: The sciences all have their own terminology and approaches and use different and often incompatible systems of measurement. Empowering science computationally involves first standardizing data and reducing complex problems to simpler, solvable ones. Our strategic intent is information at your fingertips.4 This attitude typifies the modern view of CS as the study of natural and artificial information processes. Because information processes and structures are central to much of science, there is universal value in being able to better access and manipulate them.

Health (NIH), and other large-scale funding agencies are enlightening. As Figure 1 shows, recent NSF funding trends reveal the increasing importance of computing research in various disciplines. By 2003, NSF support for math and CS had surpassed funding for all other areas, followed by physical sciences, environmental sciences, engineering, and life sciences. Among its particularly relevant priorities for 2008, the NSF included the following items: $52 million for cyber-enabled discoveries, $50 million for computing foundations, $21.5 million for science and engineering peta scale applications, and $10 million for software design and productivity. Conversations with more than 100 deans, principal researchers, and other academic leaders in 2007 revealedperhaps not surprisinglythat these academics foresee their futures as closely aligned with the NSFs computational priorities: multidisciplinary research, with computation at the core; new computer architectures, including manycore systems; large-scale data mining, analysis, and visualization; and programming in the large, concurrently, and in parallel. University recruitment priorities can offer further insight into academic perceptions of future research trends. However, whereas industry might
77

Academic Breakthroughs Guided by Funding Agencies

Academia and government are responsible for many of the most important research projects. In the US, the policies and priorities of the National Science Foundation (NSF), the National Institutes of
November/December 2009 

see new hires as a means to meet an immediate need, universities might well view them as 30-year investments and thus give priority to people who are good at reinventing themselves in response to new challenges. Furthermore, some institutions prefer to keep their hiring priorities confidential; others meet multiple priorities by making multidisciplinary hires. In general, the data suggest that academic departments hire in response to research opportunities driven by the funding organizations (and thus align themselves with the golden rulethose with the gold make the rules). For industry, it became critical to transcend existing computer architectures when excess heat production limited processor speed to approximately 4 GHz. Subsequently, researchers, architecture experts, chip designers, and manufacturers turned their attention to designing computers with multiple CPUs (cores) on a single chip. This raised many questions that the computing industry has yet to answer. At the Federated Computing Research Conference in June 2007,5 Robert Coldwell (previously of IBM Research), articulated some of these questions: Can todays operating systems and multithread applications evolve to exploit the power of many-core architectures? With per-core performance flat for the foreseeable future, what will drive consumers to purchase new systems? Given this flat performance, how can we add new features and functionalities to the runtime stack without degrading performance? Can we program safe and scalable concurrent applications with current programming languages? Are current synchronization mechanisms sufficient for creating safe and scalable applications, libraries, and runtimes? Will the existing infrastructure let us create the next generation of killer applications to make upgrading to multicore systems compelling? Can existing operating systems address energy efficiency issues by balancing throughput and power consumption? What will be the role of new technologies, such as reconfigurable fabrics, streaming machines, quantum computing, biological computing, and nanocomputing? As important as the hardware developments might be, a greater issueand one that is all too
78

Industry Looks for Breakthroughs

often ignoredis how to develop the languages and software engineering tools we need to write parallel programs. The software is particularly relevant for computational scientists because theyre the ones who must write, compile, debug, and performance tune the programs. Furthermore, computational scientists must perform these tasks routinely, not as an independent research project for each program (and preferably not with a completely new set of tools and operating system for each computer). In the past, parallel scientific computations have followed the data parallel model, with data distributed among the nodes. With future petascale machines, however, scientists must also follow a threads-parallel model in which individual processes (threads) are distributed among nodes. Theyll thus have to decide the type and number of parallel threads to exploit, which will depend on future hardware and software. Quantum computerswhich have yet to be proven feasiblewould store their data in quantum states and use superposition and entanglement to perform data operations through quantum logic gates. Because theyd store and manipulate data at the atomic level, theres tremendous potential for speed, quantity, and energy efficiency. Currently, quantum computers seem unlikely to become viable general-purpose machines; most researchers envision them as advantageous only for certain types of problems, such as factoring and discrete algorithms.6 However, government and military agencies are quite interested in decoding advanced encryptions in short times and in searching large databases; quantum computers might be applicable in such cases. Robotics is now developing in much the same way as the computer business did 30 years ago.7 Having long proven their value for industrial manipulations, robots are now appearing regularly as sophisticated toys, as standard equipment on battlefields and in operating theaters, as well as in pharmacies, in education, and in our everyday lives as care assistants, intelligent vehicles, and even vacuum cleaners. With an aging population, exploding healthcare costs, and the advantage of round-the-clock availability of robotic assistive care, the business opportunities are obvious and enormous. Microsoft Researchs investigations of robotics with academic partners have led to the Robotics Studio platform andto attract students to CSthe formation of the Institute for
Computing in Science & Engineering

Quantum Computation

Robotics

Personal Robots in Education at Georgia Tech and Bryn Mawr (http://roboteducation.org). The rapid growth of electronic games for entertainment, stimulated in part by Internet participation and special-purpose computers, is having a high marketplace impact. Although its very name can make it hard to take seriously, gamings value now extends far beyond entertainment. There are academic conferences with gaming as a major theme, and the ACM Special Interest Group for Computer Science Education (SIGCSE) now devotes a large portion of its agenda to gamingrelated activities. There are also many practical examples, including game-like simulations for training doctors, emergency responders, and other professionals. The Dartmouth Schools of Engineering and Medicine, Polytrauma, and Pandemic conferences focused on virtual medicine via gaming and simulation (http://engineering. dartmouth.edu/polytrauma), as did the Conference on Adversarial Reasoning, which targeted high-security-clearance members of the intelligence community. In addition, dozens of universities have established degree programs in gaming, and Communications of the ACM s July 2007 issue focused on creating a science of games.8 It appears timely to consider gaming technologies broader developments. Fantasy role-playing games, such as Dungeons and Dragons, are yielding spin-offs such as Second Life, which have broad appeal and are seeing growing professional use. Second Life, for example, has been used to treat Aspbergers Syndrome, while Snow World has offered pain mitigation for burn victims. Of course, educational games are well established and there appears to be growing confidence in the effectiveness of games for learning. In China, for example, where both Internet use and gaming are rising dramatically, gaming is being used for Mandarin and English language training. In the US, educational titles cover many subjects, including algebra, social awareness, discrete mathematics, symbolic logic, Arabic, French, and physical fitness. Gaming continues to affect computings direction in many ways. On the hardware side, the advent of the GUI for general computation in the 80s was hastened by graphical developments on game machines, while cell processors designed for PlayStations are now being used for numerical computations. On the theoretical side, many CS departments research the very nature of games. Some researchers, for example, view games as a model for rules of interactions governing complex,
November/December 2009 

Gaming

adaptive social-technical systems, and categorize them as either finite or infinite.9 A finite game resembles the notion of closed (terminating) computation and is played for the purpose of winning; an infinite game resembles open (nonterminating) computation and is played for the purpose of continuing the play. Current interest is moving from finite to infinite games; for example, the Web and the Internet are best viewed as infinite games, while Web data stored in databases that cant be queried by search engines is akin to a finite, rulesbased game.
Global Issues

Solving the present environmental problems will require strong research commitments. And, because a crisis is a terrible thing to waste,1 we can also look at these problems as indicators of where investments are likely to grow and where breakthroughs in computation might occur. Globally, there are many specific problematic areas: Energy: supply, efficiency, environmental impact, and the need for renewable sources. Healthcare: affordability, eradication of big killer and catastrophic diseases, adaptive technologies, and digital record keeping. Population: care for the aged, increased urbanization, and shortages in water and food supplies. Education: scale, affordability, availability, and effectiveness. Urbanization: public transportation, concentration of pollution, crime control, and the accelerated spread of disease (for the first time, according to the United Nations, 2008 saw more than half of all humans living in cities; see www. unfpa.org/swp/2007/english/introduction.html). Climate and weather: predicting important climate changes and computing weather forecasts with accuracy; to do this, we must manage huge and complex datasets and solve numerous coupled nonlinear equations (to determine, for example, whether Venice might actually be submerged before the end of this century). Another, broader area is quality of life, which is impacted by a range of factorsfrom improved communications and lifelong learning opportunities to the digitization and integration of medical records.
Multidisciplinary Developments

As problem approaches get more realistic, they often become more complex and the importance of a multidisciplinary approach grows. High-speed
79

networks, digital libraries, and virtual organizations can aid this requisite collaboration among multiple practitioners. Indeed, David Dornfeld, dean of engineering at the University of California, Berkeley, has observed that the intersection of technology, design, multidisciplinary approaches and entrepreneurship is the most important focus for the future. And yet, developments that encourage multidisciplinary approaches might have unexpected origins. For example, the digitization of data lets us link multiple datasets and subsequently reuse that data in completely unexpected ways.2 Given the golden rule of funding, much of the incentive for multidisciplinary research originates with the funding organizations, which generally encourage a team approach in order to hasten discovery and innovation. The NSF and NIH are particularly focused on exploratory research for improving computing abilities, polar and ocean research, computational neuroscience nanotechnology, education and advanced learning technologies, international collaborations, and cross-cutting efforts (for more on NSF grants, see www.nsf.gov/funding; for NIH grants, see http://grants.nih.gov/grants/funding/r21.htm). An important aspect of multidisciplinary research that discipline scientists are just now beginning to appreciate is design. In this particular context, the design approach means assembling teams with expertise in all target areas and ensuring that they all maintain a full understanding of the ultimate goals. The Technology, Entertainment, and Design (TED) conferences promote this multidisciplinary, cross-cultural perspective and maintain a vibrant site devoted to this view. Stanfords new Design Institute, which is similar to the world-famous IDEO design firm, involves six departments and aims to use design thinking to drive multidisciplinary innovation. Likewise, the Berkeley Institute of Design aims to foster a deeply holistic, multidisciplinary approach to design, spanning human-computer interaction, mechanical design, education, architecture, and art. As avant garde as these approaches might seem, design by itself does not solve problems. As Bill Buxton put it: We should not expect any silver bullets. Most of what will be invented is already here in some form. It is highly unlikely that there will be any technology that we dont know about today that will have a major impact over the next 10 or 20 years.10 Or, as the writer William Gibson put it in a 1999 National Public Radio interview, the future is already here, its just not well distributed (see www.npr.org/templates/story/ story.php?storyId=1067220).
80

A rapidly expanding multidisciplinary approach combines social sciences with computing. For example, researchers at Carnegie Mellon University are working on auction-based models for scheduling the TeraGrid, combining theoretical CS, game theory, and economics. Among the social sciences frequently involved in computational multidisciplinary work are cognitive psychology, sociology, linguistics, economics, anthropology, political science, and archaeology. Usability specialists are working with experts in artificial intelligence and robotic vision to help computers better serve us by anticipating our changing needs even before we can fully articulate them. Finally, education is a particularly important place for multidisciplinary collaborations. Georgia Techs flexible, multidisciplinary threads model combines CS instruction with additional application classes to create a four-year program tailored to each students career directions (www.cc.gatech. edu/education/undergrad/bscs). Harvey Mudd College requires computing studies for all students, while Arizona State University requires informatics training for all its freshmen. Furthermore, traditional disciplines are now assembling multidisciplinary curricula that are rich with computation. Since 2003, for example, Oregon State Universitys physics department has been offering a four-year bachelors degree in computational physics that blends five new computational physics courses with extra classes in math and CS (see www.physics. oregonstate.edu/CPUG). Also, to support the incorporation of modern computational techniques into all science education, Microsoft Research hosts faculty summits and Computational Training for Scientists workshops (http://research.microsoft. com/en-us/collaboration/focus/cs/cefs.aspx).

Research Projects of Interest

Academic research is now passing through an inflection point in response to industrys move away from faster CPUs and toward multiple CPUs on a single chip. This move is redefining research in programming languages, compilers, runtime program analysis, verification, transactional memory, operating systems, virtualization (computer resource abstraction), and speculative execution. Its also leading to a rethinking of the abstractions used to build safe, scalable concurrent applications and systems. For example, what is the optimal balance of the number of processors, of failover and backup components, and of shared and separate memories and caches when building a multiple core computer? Along those lines, University of California, Berkeley, researchers are examining the success of
Computing in Science & Engineering

parallelism at the extremes of both embedded and high-performance computing. Their aims are to make it easier to write efficient, highly parallel programs for systems with thousands of cores per chip; to explore designs rapidly with emulators based on field-programmable gate arrays (FPGAs); and to create a model independent of processor number.11 University of Washington researchers are investigating configurable systems with FPGAs and multicore processors as a means of creating high-performance computers with reconfigurable subsystems on a chip (www.ee.washington.edu/ people/faculty/hauck). On another track, Carnegie Mellon University researchers are focusing on data-intensive supercomputingthat is, on how to handle the enormous datasets arising from petascale computing (www.cs.cmu.edu/~bryant/ pubdir/cmu-cs-07-128.pdf ). In the life sciences, fundamental transformations are arising from data integration and digital libraries construction, similar to those that the Sloan Digital Sky Survey stimulated in astronomy. For example, the 4th International Workshop on Data Integration in the Life Sciences focused on helping researchers combine disparate data sets and tools and on developing semantics and controlled vocabularies for data annotation. The workshop covered projects in areas such as neuroscience, proteomics, and single nucleotide polymorphism data analysis. Other projects defy simple categorization. Symmetry studies use robotic vision to detect asymmetries in humans that might reveal difficultto-determine attributes such as lying, schizophrenia, and dangerous physical abnormalities. One researchers approach combines computer vision, computer graphics, biomedical image analysis, machine learning, intelligent robotics, and group theory to analyze images of brains, faces, and gaits, as well as for use in robotics, periodical and nearregular pattern recognition, and texture synthesis (www.cse.psu.edu/people/yanxi). Protein-folding analysis reveals complex 3D configurations that are central to developing new medicines and treatments. Computational studies in this area have made complexity more manageable and succeeded in finding important configurations that chemists had previously discarded as impossible (see http://scbmb.bcm.edu/people/ gcc_faculty_101). Virtual worlds for psychotherapy have successfully used immersive simulations to mitigate pain, and to alleviate post-traumatic stress syndrome. For example, as mentioned earlier, burn victims who are too young for pharmaceuticals but must undergo painful therapy have had their pain cut
November/December 2009 

in half when immersed in Snow Worlds virtual wintry wonderland while listening to Paul Simons music (www.hitl.washington.edu/projects/vrpain). The combination of gaming, biomedical research, graphics, and human computing is helping researchers solve complicated puzzles such as protein folding12 and the AIDS virus13 using a user-assisted optimization algorithm that leverages networked humans (www.cs.washington.edu/ homes/zoran). Programmable matter, or claytronics, describes an ensemble of mobile, intercommunicating computational units (that is, modular robots, or nanobots) that researchers can program to take arbitrary 3D shapes.14 Having such an ensemble carry out a given taskand managing them via programs running on the individual robotsis a major challenge in parallel computing.

The combination of gaming, biomedical research, graphics, and human computing is helping researchers solve complicated puzzles.

In musculoskeletal conduction for secure data communication, researchers are building humanbody-based communication technology using mobile devices; among the technologys applications are body-area-communication, user authentication, and disease diagnosis (see http://bioe.rice. edu/FacultyDetail.cfm?RiceID=95). In these investigations, the musculoskeleton system is excited with acoustic wave patterns and its response is detected with inexpensive standard mobile phones. This might be a hard password to crack! Biocomputing uses DNA and proteins to perform computational functions through biochemical, biomechanical, and bioelectronic pathways. For example, the US Department of Defense has awarded a team of nine professors from six universities $6 million over five years to exploit precise biological assembly to study quantum physics in nanoparticle arrays.15 Finally, next-generation network research takes on innovations in Internet, sensor nets, and other networks. The clean slate approach investigates a redesign of the Internet from scratch, without preconceived notions or legacy infrastructure issues.16 Although people have expressed concerns that such designs cant be validated without an experiment on a national scale, the Global Environment for Network Innovations proposed building
81

an experimental facility that would support such large-scale deployments.17

As industry and granting agencies know all too well, discovering and rewarding tomorrows innovators is difficult. By numbers alone, approximately 10 percent of all scientists write 50 percent of the scientific papers.18 This is relevant because innovators generally produce both good and bad ideas; the more investigations, the more breakthroughs. When asked by students how he got good ideas, Linus Pauling is widely reported to have said, My answer is simple: First, have a lot of ideas. Then, throw away the bad ones. This suggests that innovation would increase if we support those who write the most articles and receive the most citations. Although logical, this path tends to identify those who are already successful and well funded. Looking for projects in not-yet-fashionable research areas is useful for identifying unknown, yet talented investigators. So-called disruptive technologies are transformative precisely because they occur outside the fields current focus and are often driven by not-yet-established researchers. These researchers might not be widely published, but they can have uniquely useful insights and be effective collaborators with established teams. In fact, innovation is almost always the product of collaboration; the lone genius inventor is the exception. Were social creatures, and our creativity appears to increase with collaboration. This is especially true across multiple disciplines.19 New researchers benefit both from obtaining necessary funding and from creating a good network of relationships. One way to identify the top innovators of tomorrow is through award programs designed to attract them, such as Microsoft Researchs New Faculty Fellowship Program (http://research.microsoft. com/en-us/collaboration/awards/nff.aspx) and A. Richard Newton Breakthrough Research Awards. Although the awards are modest when compared to funding from national agencies, theyre overheadfree and focused on producing high-impact results in risky areas. Government funding agencies are also investing in riskier, exploratory research. In August 2008, the NSFs Directorate for Computer and Information Science and Engineering established four new Expeditions in Computing awards of $10 million each to allow teams of researchers and educators to pursue far-reaching agendas that promise significant advances in computing.20 As much as we might try to grow it, innovation often appears where its least expected. As General
82

How to Recognize Tomorrows Innovators

Motors head of product development notes, investment and enthusiasm for clean technologies in Asia is so great that cars powered by fuel cells are likely to take off in China before they do in the United States.21 In general, developing countries tend to have high levels of early stage entrepreneurship, more new ventures, and greater necessity. As production cycles become increasingly shorter, executives at companies such as 3M and Procter & Gamble believe that they must innovate faster just to maintain their current market share. Procter & Gamble mapped consumer goods life cycles over 10 years and found that theyd fallen by half. In turn, 3M, which is well known for innovation, tries to shorten development times by outsourcing innovative activities the company previously performed in-house (concurrent development) and by communicating with customers much earlier in the product life cycle. Its important for managers, funders, and researchers to appreciate that breakthroughs take time. Innovations such as lasers, high-speed electronics, microwave ovens, and the Web appeared decades after the basic research that spun them off. Examples in the pharmaceutical field include penicillin, Rogaine, and Viagra. In applied, problem-oriented research, it typically takes three years before results emerge. Also, the law of unintended consequences states that, for every six outcomes, one unpredictable result will occur.22 Thus, we should learn to expect the unexpected. ext-generation innovation and breakthrough research appear inextricably intertwined with the future of computing. While humans excel at understanding and inventing, leveraging computational power strongly complements our ingenuity and increases the impact of our intellects. Although we look forward to and should cherish the next Einstein, Babbage, and Feynman, relying on the lone genius working in a backroom to achieve breakthroughs is an ineffective model for science planning.23 Geniuses will do their own thing in their own space because they must; meanwhile, the majority of scientists could simply use some assistance. Assembling teams designed for creativity enhances the likelihood of innovation, sometimes via the unforeseen intersections of bright ideas and sometimes due to our social and competitive natures. To create an effective team, it appears necessary to encourage a diversity of expertise, approaches, and cultural perspectives;
Computing in Science & Engineering

 engage individuals open to multidisciplinary thinking; generate as many new ideas as quickly as possible; adopt a tolerance for failure; and employ collaboration tools that enable rapid group problem solving. The idea that progress in research proceeds via breakthroughs probably has some truth to it. However, its only part of a much larger story. Popular culture and history books often focus on the moment of insight, but breakthroughs are often side products of many years of curiosity-driven research. Consequently, to increase innovation, we should increase funding, develop patience for long research efforts, invest in new and unproven researchers, and explore little-known areas. The more research we do, the higher the breakthrough probability and the greater the opportunity for transformative innovation.

12. J. Bohannon, Gamers Unravel the Secret Life of

13.

14.

15.

16.

17.

References
1. V. Khosla, A Special Report on Innovation, The

Economist, Oct. 2007, p. 3.

2. S. Vincent-Lancrin, What Is Changing in Academic

18.

3. 4. 5.

6.

7. 8.

9. 10.

11.

Research? Trends and Futures Scenarios, European J. Education, vol. 41, no. 2, 2006, pp. 169202. N.N. Taleb, The Black Swan: The Impact of the Highly Improbable, Random House, 2007. P.J. Denning, Computing Is a Natural Science, Comm. ACM, vol. 50, no. 14, 2007, pp. 1318. R. Colwell, Computer Architecture Futures 2007, PowerPoint presentation, Federated Computing Research Conf., 2007; http://lazowska.cs.washington. edu/fcrc/Colwell.FCRC.pdf. I. Glendenning, The Bloch Sphere, presentation at the European Centre for Parallel Computing at Vienna, Feb. 2005; www.vcpc.univie.ac.at/~ian/ hotlist/qc/talks/bloch-sphere.pdf. W.G. Gates, A Robot in Every Home, Scientific Am., vol. 296, no. 1, 2007, pp. 4451. M. Zyda, ed., special issue, Creating a Science of Games, Comm. ACM, vol. 50, no. 7, 2007; http:// portal.acm.org/toc.cfm?id=1272516&type=issue. J.P. Carse, Finite and Infinite Games, Ballantine, 1986, p. 1; www.glg.net/pdf/Finite_Infinite_Games.pdf. W. Buxton, Sketching User Experience: Getting the Design Right and the Right Design, Elsevier, 2007. K. Asanovic et al., The Landscape of Parallel Computing Research: A View from Berkeley, Dec. 2006; www.eecs.berkeley.edu/Pubs/TechRpts/2006/ EECS-2006-183.pdf.

19.

20.

21.

22.

23.

Protein, Wired, vol. 17, no. 7, 2009; www.wired. com/medtech/genetics/magazine/17-05/ff_ protein. A. Dworkin, Could This Video Game Cure AIDS? The Oregonian, 8 May 2008; http://blog.oregonlive. com/pulse/2008/05/could_this_video_game_cure_ aid.html. S. Goldstein, J. Campbell, and T. Mowry, Programmable Matter, Computer, vol. 38, no. 6, 2005, pp. 99101. Biologically Based Quantum Computers? DNA, Proteins and Peptides Could Help Construct New Nanoscale Electronics, Science Daily, 21Mar. 2007; www.sciencedaily.com/releases/ 2007/03/070321093451.htm. N. Dukkipati et al., Congestion Control in 100x100: Why TCP is a Poor Choice and How to Redesign It from Scratch, tech. report, High Performance Networking Group, Stanford Univ., Dec. 2004; http://yuba. stanford.edu/rcp/100x100Dec04.pdf. S. Shenker, We Dream of GENI: Exploring Radical Network Designs, presentation, CRA Computing Community Consortium, 2007; http://lazowska. cs.washington.edu/fcrc/Shenker.FCRC.pdf. D.K. Simonton, Creative Productivity: A Predictive and Explanatory Model of Career Trajectories and Landmarks, Psychological Rev., vol. 104, no. 1, 1997, pp. 6689. F. Johansson, The Medici Effect: What Elephants and Epidemics Can Teach Us About Innovation, Harvard Business School Press, 2006. US Natl Science Foundation, NSF Announced Expeditions in Computing Awards, 18 Aug. 2008, www.nsf.gov/news/news_summ.jsp?cntn_ id=112075&govDel=USNSF_51. V. Vaitheeswaran, Something New Under the Sun: A Special Report on Innovation, The Economist, 13 Oct. 2007, pp. 323. R. Norton, Unintended Consequences, The Concise Encyclopedia of Economics, 2nd ed., Library of Economics and Liberty, 2007; www.econlib.org/ library/Enc/UnintendedConsequences.html. R.K. Sawyer, Group Genius: the Creative Power of Collaborations, Basic Books, 2007, p. 7.

Thomas C. McMail is senior research program manager at Microsoft Research, where he is responsible for research outreach programs in North America. Contact him at [email protected].

Selected articles and columns from IEEE Computer Society publications are also available for free at http://ComputingNow.computer.org.

November/December 2009 

83