October 16, 2002
MIT Media Labs
2:00-3:30 The Present
Neil Gershenfeld: Bits and atoms
John Doyle (Caltech): Engineering complexity
Walter Willinger (AT&T): The Internet
David Chassin (PNNL): The power grid (video)
Brian Bramlett (Intel): Integrated circuits
Noubar Afeyan (Flagship Ventures/Affinnova): Evolutionary product design
Mark Ptashne (Sloan-Kettering): Biological evolution
Bill Butera: Paintable computing
Saul Griffith: Predicated self-assembly
Raffi Krikorian: Internet 0
Josh Lifton: Pushpin computing
Kamal Malek: Interactive design by evolutionary algorithms
Yanni Loukissa: Generative fabrication
Russell Tedrake: Reinforcement learning for dynamically stable legged locomotion
4:45-6:15 The Future
Raff D'Andrea (Cornell): Control of complex systems
Cynthia Breazeal: Biologically inspired design for (more) scalable robots
Joe Jacobson: Self-replicating systems
Ike Chuang: Threshold for life
Kamal Abdali (NSF)
Sri Kumar (DARPA)
6:15-8:00 Open White-Board Reception
CBA research projects, ranging from printing active electronics to painting computers to programming molecular machines, promise to enable the creation of engineered systems on an unprecedented scale. But if these are designed based on current engineering practice, the only thing that can be said with confidence about them is that they are likely to not work. As systems pass from having millions to billions to trillions of components, and beyond, a recurring lesson has been that they encounter unanticipated emergent failure mechanisms. The serious prospect of "Avogadro-scale" engineering is going to require the development of a real theory of engineering emergence that can guide the creation of enormously complex systems without explicitly specifying how they work, so that success rather than failure can be an emergent property.
Unfortunately, work in this area has tended to fall into one of two disjoints sets: deep thinking about fundamental scaling principles that are not reduced to practice, and small demonstrations that do not scale to large sizes. The casual invocation of terms like "complex," "nonlinear," and "emergent" has served to obfuscate rather than clarify their meaning, neglecting the importance of both mathematical rigor and domain-specific application details. It is this relatively neglected intersection between fundamental theory and applied practice in emergent systems that we hope to explore in this workshop, which reprises an influential series of meetings that were held in Caltech in 1997. We've gathered experts on scaling in the Internet, power grid, chip design, product engineering, robotics, and molecular biology to speak on the relevant issues and opportunities in their domains, and then will look ahead to the consequences for our understanding of self-reproducing systems and ultimately life itself.
Taken together, these topics ask rather than answer important questions: Is there a deeper story that transcends the particular features of the anecdotal examples? Can attributes such as hierarchy, adaptation, and evolution be designed with the same rigor we now bring to understanding the role of bandwidth, power, or noise? What are the implications of this possibility for the organization of inquiry around it?
We believe that we do see the first stirrings of a predictive theory that can guide scientific investigation as well as engineering practice in designing systems with an enormous number of interacting degrees of freedom, but also see its development demanding a disciplined approach to interdisciplinary research that can transcend the traditional tension between rigor and relevance. We look forward to testing these beliefs with you at this workshop, and beyond.
October 16, 2002
S. Kamal Abdali is Acting Division Director of the Computer-Communications Research Division at NSF. He received a Ph.D. in Computer Science from the University of Wisconsin, Madison, in 1974. He has been a computer science faculty member in New York University and Rensselaer Polytechnic Institute, and has held adjunct appointments at Oregon Graduate Institute and the University of Delaware. Prior to joining NSF, he was a principal scientist at the computer research lab in Tektronix, and led the symbolic computation research group there.
His research has spanned the combinatory and lambda calculi, programming language semantics, and computer algebra language and systems design. His current interests include symbolic and algebraic computation, computer algebra systems, and automated theorem proving.
Noubar B. Afeyan, Ph.D., Senior Managing Director and CEO of Flagship Ventures, is a recognized technologist and entrepreneur. Prior to the formation of Flagship Ventures, he was President and CEO of NewcoGen, a company he founded in 1999. In addition, he was the Managing General Partner of AGTC, a $150 million venture capital fund for the genomics industry, and he was a General Partner at OneLiberty. Dr. Afeyan is a frequent guest speaker at technology forums throughout the country and is a Senior Lecturer at MIT's Sloan School of Management.
Until 1999, Dr. Afeyan was Senior Vice President and Chief Business Officer of Applera Corporation (previously PE Corp) (NYSE: ABI). While at Applera, he initiated and oversaw the creation of Celera Genomics (NYSE: CRA), a tracking stock subsidiary of Applera Corporation focused on generating and providing genomic information to the pharmaceutical and biomedical research industries. Celera Genomics grew from $0.5 billion valuation to over $10 billion in just nine months after it was brought to market in 1999.
Previously, Dr. Afeyan was the Founder, Chairman, and CEO of PerSeptive Biosystems (formerly Nasdaq: PBIO), a leader in the bio-instrumentation field, where he oversaw PerSeptive's growth from $1 million in revenues in 1991 to $100 million in revenues in 1997. PerSeptive merged with PE Corporation in 1998 in a deal valued at $360 million. During 1996 and 1997, Dr. Afeyan also served as Chairman of the Board of ChemGenics Pharmaceuticals, a privately held genomics and drug discovery company spun out of PerSeptive Biosystems. ChemGenics was acquired by Millennium Pharmaceuticals (Nasdaq: MLNM) in 1997 in a deal valued at $100 million.
Dr. Afeyan has also been a founding investor and active board member/advisor for several other high-tech startups including Antigenics (Nasdaq: AGEN), Color Kinetics, and EXACT Sciences (Nasdaq: EXAS). He currently serves as a Director for Antigenics and Color Kinetics, and is a member of the Board of Governors of Boston University Medical School, the Board of Associates for the Whitehead Institute at MIT, the Advisory Council of the McGowan Institute for Regenerative Medicine at the University of Pittsburgh, the Advisory Board of the Center for Bits and Atoms at MIT's Media Lab, the Advisory Board of the Faculty of Engineering at McGill University, and the Board of the MIT Enterprise Forum. He has authored numerous publications and patents. Dr. Afeyan earned his undergraduate degree in Chemical Engineering from McGill University in Montreal and his Ph.D. in Biochemical Engineering from the Massachusetts Institute of Technology ("MIT").
Brian Bramlett (BS Appl. Phys '89, B EE '90, MSEE '92 Georgia Institute of Technology) has been with Intel since 1992 in roles including microprocessor design (Pentium® ProTM through Pentium® 4), EDA CAD (Logic Verification and Validation, Physical Design, Frameworks), Internet technologies (creation of www.intel.com, Internet-television infrastructure with involvement in Mozilla Open Source development), and Consumer Electronics/Peripherals (IntelPlayTM and recent patent work).
He currently works in desktop microprocessor design with roles in System Engineering and Process Improvement, and is pursuing PhD research in this field. Prior to joining Intel, Brian was responsible for design, deployment, and initial operation of the Electron Beam Nanolithography capability at the Georgia Tech Microelectronics Research Center with related research in fabrication and device physics at Tektronix Solid State Research Labs and Georgia Tech. Brian is a member of IEEE and ACM.
Cynthia Breazeal is an assistant professor of Media Arts and Sciences at the MIT Media Lab. She is Director of the Robotic Life Group and holds the LG Group career development chair.
Cynthia has been building autonomous robots for over a decade ranging from insect-like planetary micro-rovers, to upper-torso humanoids, to expressive anthropomorphic faces, and more. Her work is informed by scientific theories of natural behavior and incorporates artistic insights to create capable and appealing robot creatures that can engage us, communicate with us, and learn from us on our terms. Her current research extends these themes in the area of human-robot relations, whether the robot is like a creature that you interact with socially, an avatar that you project yourself into to interact with others across the world, or a robot that you wear to augment your physical abilities.
She has published extensively in journals, conference proceedings, books, and magazines. Her most recent book, Designing Sociable Robots, is published by The MIT Press (2002), and she is co-editor of Biologically Inspired Intelligent Robots, published by SPIE Press (forthcoming). She has been featured in Time magazine as an inventor, and recognized as a prominent young innovator (Boston Business Forward, and Richard Saul Wurman's 1000). She served as an expert consultant for the Spielberg/Kubrick movie Artificial Intelligence. Her work has been nationally and internationally featured in the media including the NBC Nightly News, ABC World News Tonight, Business Week, US News and World Reports, Scientific American, The London Times, Newton, Le Figaro, The New York Times, The Washington Post, NPR's Morning Edition, and more.
She received her B.S. (1989) in Electrical and Computer Engineering from the University of California at Santa Barbara. She received her S.M (1993) and Sc.D. (2000) in Electrical Engineering and Computer Science from the Massachusetts Institute of Technology. Her graduate and postdoctoral research was carried out at the MIT Artificial Intelligence Lab. While at the AI Lab, she participated in the development of some of the world's most famous robots including the upper torso humanoid robot, Cog, and the sociable robot, Kismet.
David P. Chassin. Dave Chassin’s innovative leadership has resulted in advances in the basic technologies needed to develop more comfortable, efficient, and cost-effective buildings. This includes projects to enable grid-friendly systems, the widespread deployment of advanced equipment diagnostics, and the use of wireless data collection and control systems in buildings. Dave has 15 years of experience in the research and development of computer software applications for the architecture, engineering, and construction
His expertise encompasses:
Dave leads a project team to develop low-cost distributed load control and grid-friendly chips for building systems. These 2- x 2.5-inch circuit boards, installed in appliances such as air conditioners and water heaters, monitor the power grid and turn off for a short period of time in response to grid overload. By improving demand response in the event of a crisis, these inexpensive chips will result in more reliable power grids, smaller electricity bills, and a cleaner environment. Dave served as technical lead in the development of the Whole-Building Diagnostician (WBD), a set of real-time building energy diagnosticians that detect poorly operating building energy systems. The WBD is deployed in more than 300 buildings around the United States. He also was technical manager for the development of Softdesk Energy, a building energy analysis and simulation package for AutoCAD software that includes engineering drawing interpretation.
Dave was chief engineer, product designer, and project manager for the development of CAD Overlay, a family of hybrid raster/vector processing and editing products for AutoCAD. Dave also conducts research on system dynamics and complex system modeling and simulation. Some of his recent work has led to the discovery of an abstract machine that may be useful in establishing a robust econophysical model of very large scale market-based power systems controls.
Isaac Chuang is an Associate Professor at the Massachusetts Institute of Technology. He leads the quanta research group at the Center for Bits and Atoms located at the MIT Media Laboratory. This group seeks to understand and create information technology and intelligence from the fundamental building blocks of physical media, atoms and molecules. Our research interests include:
Chuang is co-author of a textbook on Quantum Computation and Quantum Information. He is also on the editorial board of theVirtual Journal of Quantum Information. He is a graduate of Stanford University, where he was a Hertz Foundation Fellow, and of MIT.
Raffaello D'Andrea obtained a Ph.D. in electrical engineering from the California Institute of Technology in 1997. Since then, he has been with the Department of Mechanical and Aerospace Engineering at Cornell University, where he is an Associate Professor. He is also a member of the Applied Mathematics, Electrical Engineering, and Theoretical and Applied Mechanics fields at Cornell. His research and teaching interests include the development and application of tools for controlling complex systems.
Dr. D'Andrea has been the recipient of the American Control Council O. Hugo Schuck Best Paper award, an NSF Career Award, a DOD sponsored Presidential Early Career Award for Scientists and Engineers (PECASE), and was a Distinguished Lecturer in the NSF Research Highlight Series. He was the system architect and faculty advisor for the world champion Cornell Autonomous Robot Soccer team in 1999 (Stockholm, Sweden), 2000 (Melbourne, Australia), and 2002 (Fukuoka, Japan). He has appeared with the Cornell RoboCup team on Scientific American Frontiers (1999 and 2001), the Lemelson Center at the Smithsonian (1999), and the Tech Museum of Innovation in San Jose (2001). His recent collaboration with Canadian artist Max Dean "The Table: Childhood", an interactive piece of art, was on display at the Biennale di Venezia, the world's most prestigious contemporary art exhibit, and will be on exhibit at the National Gallery of Canada in 2003.
John Doyleis Professor of Control and Dynamical Systems, Bioengineering, and Electrical Engineering and at the California Institute of Technology. He has a BS and MS in EE, from MIT, 1977 and a PhD in mathematics, UC-Berkeley, 1984. His current research interests are in theoretical foundations for complex networks, primarily in engineering and biology, and the interplay between robustness, feedback, control, dynamical systems, computation, communications, and statistical physics. Additional interests include theoretical foundations of multiscale physics and financial markets. Prize papers include the IEEE Baker (also ranked in the top 10 ``most important'' papers world-wide in pure and applied mathematics from 1981-1993), the IEEE AC Transactions Axelby (twice), and the AACC Schuck. Individual awards include the IEEE Centennial Outstanding Young Engineer, the IEEE Hickernell, the American Automatic Control Council (AACC) Eckman, and the Bernard Friedman. He has held national and world records and championships in various sports.
Neil Gershenfeld is the Director of MIT's Center for Bits and Atoms, an interdisciplinary initiative that is broadly exploring how the content of information relates to its physical representation, from atomic nuclei to global networks. CBA's intellectual community and research resources cut across traditional divisions of inquiry by disciplines and length scales in order to bring together the best features of the bits of new digital worlds with the atoms of the physical world. Dr. Gershenfeld has also led the Media Lab's Things That Think industrial research consortium, which pioneered moving computation out of conventional computers and into the rest of the world, and works with the Media Lab Asia to help coordinate the technical guidance for this ambitious international effort based in India that is investigating technology for global development.
His own laboratory studies fundamental mechanisms for manipulating information (which led to the development of molecular logic used to implement one of the first complete quantum computations and to analog circuits that can efficiently perform optimal digital operations), the integration of these ideas into everyday objects such as furniture (seen in the Museum of Modern Art and used in automobile safety systems), and applications with partners ranging from developing a computerized cello for Yo-Yo Ma and stage for the Flying Karamazov Brothers to instrumentation used by rural Indian villagers and nomadic reindeer herders.
Beyond his many technical publications and patents, he is the author of best-selling books including "When Things Start To Think" and the texts "The Nature of Mathematical Modeling" and "The Physics of Information Technology." His work has been featured by the White House and Smithsonian Institution in their Millennium celebrations, and been the subject of print, radio, and TV programs in media including the New York Times, The Economist, CNN, and PBS.
Dr. Gershenfeld has a B.A. in Physics with High Honors from Swarthmore College, was a member of the research staff at Bell Labs where he studied laser interactions with atomic and nuclear systems, received a Ph.D. in Applied Physics from Cornell University for experimental tests of order in complex condensed matter systems, and was a Junior Fellow of the Harvard Society of Fellows where he ran an international study on prediction techniques.
Joe Jacobson is Associate Professor at the Massachusetts Institute of Technology where he is co-founder and co-PI of the Center for Bits and Atoms and leads the Molecular Machine Group which has pioneered research in logic and machines developed from inorganic and biochemical molecular building blocks. Jacobson was educated at Brown, MIT (Ph.D., Physics) and Stanford (ERATO Post-Doctoral Fellow, Nonlinear Quantum Structures) and is the recipient of the 1999 TR100 Award for Innovation, The 2000 Gutenberg Prize, and the 2001 Discover Award.
Sri (Srikanta) Kumar is a Program Manager at DARPA, Information Processing Technology Office. His current programs include Bio-Computation, Network Modeling and Simulation, and Sensor Information Technology. He is also Senior Technical Advisor in the Information Technology Laoratory at NIST, from which he is on detail to DARPA. Before joing the government in 1998, he served on the faculty of Northwestern University (85-98), Electrical and Computer Engineering Department, where he was also the founding director of the Executive Masters Program in Information Technology. During 82-85, he was on the faculty of ECSE Department at RPI. He received his Ph. D from Yale, in Engineering and Applied Science.
Mark Ptashne is the Ludwig Professor of Molecular Biology at the Sloan Kettering Institute in New York. He received his Ph.D. from Harvard University in Cambridge, Mass. and was a faculty member there from 1970 to 1997. He has studied how proteins bind DNA and activate or repress transcription, beginning with phage lambda and then turning to eukaryotes, particularly yeast. He has written two books on this subject (A Genetic Switch, 1992, Cell Press and Blackwell Scientific Publications and, with Alex Gann, Genes and Signals, 2002, Cold Spring Harbor Laboratory Press). Ptashne received the Lasker award for Basic Research in 1997, as well as the Gairdner, Horwitz, General Motors Sloan Foundation, and Charles-Leopold Mayer (French Academie des Sciences) Prizes. He's a member of the National Academy of Sciences, was a Junior Fellow of the Harvard Society of Fellows, founded the Genetics Institute biotech company (and practices the violin 2-3 hours a day).
Walter Willinger received the Diplom (Dipl. Math.) from the ETH Zurich, Switzerland, and the M.S. and Ph.D. degrees from the School of ORIE, Cornell University, Ithaca, NY, and is currently a member of the Information and Software Systems Research Center at AT&T Labs - Research, Florham Park, NJ. Before that, he worked from 1986-1996 at Bellcore (now Telcordia Technologies). He has been a leader of the work on the self-similar (``fractal'') nature of data network traffic and is interested in understanding the behavior of large-scale complex communication networks such as the Internet.
The Paintable Computing project is developing both hardware and software for computing on huge ensembles of millimeter-scale "sand-sized" computing nodes. The goal is to create reliable applications that run on thousands of nodes, yet to treat those nodes with the abandon common to bulk materials. We demonstrate work on both hardware and software. A device simulator shows tools and applications that are built on mobile code-modules that self-assemble into complex software hierarchies, suitable for applications. Hardware development is continuing with the work on pushpin computers. Here, we show device prototypes in various stages of development -- all works in progress. Please come see us in the spring!
Biology uses many examples of programmed, and self-assembly processes to achieve a remarkable array of structure and morphologies at many scales. Whilst our understanding of self-assembling processes is slowly proceeding, there are no non-biological examples of on-the-fly programmable assembly processes for three dimensional structure at any scale. Sub-cellular, uni-cellular, and multi-cellular biological systems all display controlled programming of 3D structure. Allosteric and catalytic assembly will be demonstrated in mechanical self-assembling systems as a step towards 'predicated assembly'. Adding 'state' to self-assembling sub-units in this way is one route towards model mechanically self-replicating systems we have designed. Early studies in 1D-3D programmatic folding systems will be shown introducing a concept of '3D-completeness'.
I0 is an infrastructure for networking large numbers of small devices. Traditional Internet implementations are simply too complicated for micro devices to understand and use -- we are working towards creating new standardized physical and logical layers for networking that allows small computationally restricted devices to self organize and cooperate while still remaining fully functional with the Internet of today. Our devices, in about 2K of code, are fully functional Internet nodes that can be distributed and embedded in buildings and sensor networks.
Pushpin Computing is a testbed of 100 independent sensor nodes equipped with onboard processing and communication capabilities. We use this platform in conjunction with a novel programming model as a testbed for realizing and testing ideas pertaining to distributed sensor networks and networks in general.
Interactive Design by Evolutionary Algorithn
Kevin Karty, Kamal Malek, and Matt Utterback (Affinnova)
Interactive Design by Evolutionary Algorithn, or IDEATM is Affinnova's web-enabled system for evolving product designs under consumer feedback, in real time. IDEATM is based on a multi-user interactive genetic algorithm, designed to maintain population diversity. The system is thus able concurrently to evolve the most preferred designs, from a very large universe of possibilities, for several market segments.
Yanni Loukissas and Larry Sass
Architects typically create physical models of their designs in order to understand the building’s geometry and size. Traditionally these physical models, typically presented to a client, are made from an architect’s conceptual and finished drawings. Here we are also creating physical models using rapid prototyping machines from CAD files, versus drawings and the files are generated from shape rules created in CAD, not traditional architectural rules used to create most buildings. The two goals of the research are to create complex shapes using generative programs written as part of larger CAD systems, and second to visualize complex shapes by printing them. The next step in the research will be to construct larger scale physical mockups from these small scale files to see if it is possible to generate real building details from shape rules and new found
Reinforcement Learning for Dynamically Stable Legged Locomotion
Russ Tedrake and H. Sebastian Seung
The limit cycle stability of walking robots with only two legs, and even quickly moving robots with four or more legs, cannot be described as transitions between a set of stable fixed points in the configuration space of the robot. Our ability to quantify and generate this dynamic stability has been limited by the complex dynamics of our robots, the fact that many of them are underactuated, and the uncertainty involved with walking on an unmodeled surface. My approach is to formulate the problem of dynamic stability as a reinforcement learning problem with delayed reward. I'll demonstrate preliminary, but promising, results using a 3D simulation of a simple legged robot.