Letters to the Editor
Total Page:16
File Type:pdf, Size:1020Kb
Articles Letters to the Editor Research Priorities for is a product of human intelligence; we puter scientists, innovators, entrepre - cannot predict what we might achieve neurs, statisti cians, journalists, engi - Robust and Beneficial when this intelligence is magnified by neers, authors, professors, teachers, stu - Artificial Intelligence: the tools AI may provide, but the eradi - dents, CEOs, economists, developers, An Open Letter cation of disease and poverty are not philosophers, artists, futurists, physi - unfathomable. Because of the great cists, filmmakers, health-care profes - rtificial intelligence (AI) research potential of AI, it is important to sionals, research analysts, and members Ahas explored a variety of problems research how to reap its benefits while of many other fields. The earliest signa - and approaches since its inception, but avoiding potential pitfalls. tories follow, reproduced in order and as for the last 20 years or so has been The progress in AI research makes it they signed. For the complete list, see focused on the problems surrounding timely to focus research not only on tinyurl.com/ailetter. - ed. the construction of intelligent agents making AI more capable, but also on Stuart Russell, Berkeley, Professor of Com - — systems that perceive and act in maximizing the societal benefit of AI. puter Science, director of the Center for some environment. In this context, Such considerations motivated the “intelligence” is related to statistical Intelligent Systems, and coauthor of the AAAI 2008–09 Presidential Panel on standard textbook Artificial Intelligence: a and economic notions of rationality — Long-Term AI Futures and other proj - Modern Approach colloquially, the ability to make good ects on AI impacts, and constitute a sig - Tom Dietterich, Oregon State, President of decisions, plans, or inferences. The nificant expansion of the field of AI AAAI, Professor and Director of Intelligent adoption of probabilistic and decision- itself, which up to now has focused Systems theoretic representations and statistical largely on techniques that are neutral Eric Horvitz , Microsoft research director, ex learning methods has led to a large with respect to purpose. We recom - AAAI president, cochair of the AAAI presi - degree of integration and cross-fertil - mend expanded research aimed at dential panel on long-term AI futures ization among AI, machine learning, ensuring that increasingly capable AI Bart Selman, Cornell, Professor of Computer statistics, control theory, neuroscience, systems are robust and beneficial: our AI Science, cochair of the AAAI presidential and other fields. The establishment of systems must do what we want them to panel on long-term AI futures shared theoretical frameworks, com - do. The attached research priorities doc - Francesca Rossi, Padova & Harvard, Professor bined with the availability of data and ument [see page 105 of this issue of AI of Computer Science, IJCAI President and processing power, has yielded remark - Magazine ] gives many examples of such Cochair of AAAI Committee on Impact of able successes in various component research directions that can help maxi - AI and Ethical Issues tasks such as speech recognition, image mize the societal benefit of AI. This Demis Hassabis, cofounder of DeepMind classification, auton omous vehicles, research is by necessity interdiscipli - Shane Legg, cofounder of DeepMind machine translation, legged locomo - nary, because it involves both society Mustafa Suleyman, cofounder of DeepMind tion, and question-answering systems. and AI. It ranges from economics, law, As capabilities in these areas and Dileep George, cofounder of Vicarious and philosophy to computer security, Scott Phoenix, cofounder of Vicarious others cross the threshold from labora - formal methods, and, of course, various Yann LeCun, head of Facebook’s Artificial tory research to economically valuable branches of AI itself. Intelligence Laboratory technologies, a virtuous cycle takes In summary, we believe that research Geoffrey Hinton, University of Toronto and hold whereby even small improve - on how to make AI systems robust and Google Inc. ments in performance are worth large beneficial is both important and timely, sums of money, prompting greater Yoshua Bengio, Université de Montréal and that there are concrete research investments in research. There is now a directions that can be pursued today. Peter Norvig, Director of research at Google broad consensus that AI research is and coauthor of the standard textbook Arti - progressing steadily, and that its Editor’s Note: This letter has been ficial Intelligence: a Modern Approach impact on society is likely to increase. signed, by press time, by nearly 7,000 Oren Etzioni, CEO of Allen Inst. for AI The potential benefits are huge, since persons, including many AAAI Fellows. Guruduth Banavar, VP, Cognitive Comput - everything that civilization has to offer Signatories des cribe themselves as com - ing, IBM Research Copyright © 2015, Association for the Advancement of Artificial Intelligence. All rights reserved. ISSN 0738-4602 WINTER 2015 3 Articles Michael Wooldridge, Oxford, Head of Dept. Machine Intelligence Research Institute George Dvorsky, Contributing Editor, io9; of Computer Science, Chair of European (MIRI) Chair of the Board, Institute for Ethics and Coordinating Committee for Artificial Intel - Eliezer Yudkowsky, MIRI researcher, cofounder Emerging Technologies ligence of MIRI (then known as SIAI) George Church, Harvard & MIT Leslie Pack Kaelbling, MIT, Professor of Com - Katja Grace, MIRI researcher Klaus-Dieter Althoff, University of puter Science and Engineering, founder of Benja Fallenstein, MIRI researcher Hildesheim, Professor of Artificial Intelli - the Journal of Machine Learning Research Nate Soares, MIRI researcher gence; Head of Competence Center Case- Tom Mitchell, CMU, former President of Based Reasoning, German Research Center Paul Christiano, Berkeley, Computer Science AAAI, chair of Machine Learning Depart - for Artificial Intelligence, Kaiserslautern; graduate student ment Editor-in-Chief German Journal on Artificial Anders Sandberg, Oxford, Future of Human - Toby Walsh, Univ. of New South Wales & Intelligence ity Institute researcher (Oxford Martin NICTA, Professor of AI and President of the Christopher Bishop, Distinguished Scientist, School) AI Access Foundation Microsoft Research Daniel Dewey, Oxford, Future of Humanity Murray Shanahan, Imperial College, Profes - Vernor Vinge, San Diego, Professor Emeritus Institute researcher (Oxford Martin School) sor of Cognitive Robotics of Computer Science Stuart Armstrong, Oxford, Future of Human - Michael Osborne, Oxford, Associate Professor ity Institute researcher (Oxford Martin Steve Crossan, Google of Machine Learning School) Charina Choi, Google David Parkes, Harvard, Professor of Com - Toby Ord, Oxford, Future of Humanity Insti - Matthew Putman, CEO of Nanotronics Imag - puter Science tute researcher (Oxford Martin School), ing Laurent Orseau, Google DeepMind Founder of Giving What We Can Owain Evans, MIT, Ph.D. student in proba - Ilya Sutskever, Google, AI researcher Neil Jacobstein, Singularity University bilistic computing Blaise Aguera y Arcas, Google, AI researcher Dominik Grewe, Google DeepMind Viktoriya Krakovna, Harvard, Statistics Ph.D. Joscha Bach, MIT, AI researcher Roman V. Yampolskiy, University of student, FLI cofounder Bill Hibbard, Madison, AI researcher Louisville Janos Kramar, FLI researcher Steve Omohundro, AI researcher Vincent C. Müller, ACT/Anatolia College Ryan Calo, U. Washington, Assistant Profes - Ben Goertzel, OpenCog Foundation Amnon H Eden, University Essex sor of Law Richard Mallah, Cambridge Semantics, Henry Kautz, University of Rochester Heather Roff Perkins, U. Denver, visiting pro - fessor Director of Advanced Analytics, AI re - Boris Debic, Google, Chief History Officer Tomaso Poggio, Director, Center for Brains, search er Kevin Leyton-Brown, University of British Alexander Wissner-Gross, Harvard, Fellow at Columbia, Professor of Computer Science Minds and Machines the Institute for Applied Computational Sci - Trevor Back, Google DeepMind Joshua Greene, Harvard, Associate Professor of Psychology ence Moshe Vardi, Rice University, editor-in-chief Adrian Weller, Cambridge, AI researcher of Communications of the ACM Anthony Aguirre, Santa Cruz, Professor of Physics, cofounder of FLI Jacob Steinhardt, Stanford, AI Ph.D. student Peter Sincak, prof. TU Kosice, Slovakia Frank Wilczek, MIT, Professor of Physics, Nick Hay, Berkeley, AI Ph.D. student Tom Schaul, Google DeepMind Nobel Laureate for his work on the strong Jaan Tallinn, cofounder of Skype, CSER, and Grady Booch, IBM Fellow nuclear force FLI Alan Mackworth, Professor of Computer Sci - Marin Soljacic, MIT, Professor of Physics, Elon Musk, SpaceX, Tesla Motors ence, University of British Columbia. Ex McArthur Fellow, Founder of WiTricity Steve Wozniak, cofounder of Apple AAAI President Max Tegmark, MIT, Professor of Physics, Luke Nosek, Founders Fund Andrew Davison, Professor of Robot Vision, cofounder of FLI and FQXi Director of the Dyson Robotics Lab at Impe - Aaron VanDevender, Founders Fund Meia Chita-Tegmark, Boston University, rial College London Erik Brynjolfsson, MIT, Professor at and direc - cofounder of FLI Daniel Weld, WRF / TJ Cable Professor of tor of MIT Initiative on the Digital Economy Michael Vassar, founder of MetaMed and ex- Computer Science & Engineering, Universi - Margaret Boden, U. Sussex, Professor of Cog - ty of Washington