全球808名专家1980名业界领袖联署23条AI军规

【新智元导读】1月初，在名为Benificial AI 2017的会议上，近2000名产学研领袖和中坚力量共同联名发布“AI 23条原则”，从研究问题、伦理价值观和长期问题三个角度宏观地为AI发展提出规定。联名者包括：Demis Hassabis，Ilya Sutskeve，Yann LeCun，Yoshua Bengio以及霍金和马斯克等等。面对AI可能会被误用和威胁人类的担忧，这23条原则被期待能确保AI走在“正轨”，以获得更加健康的发展。

1月5日到8日举行的Benificial AI 2017会议上，1980名来自产业、学术和研究界的AI相关人士共同制定“23条原则”，确保AI的健康发展。

先来看下部分签名的大牛们：

Demis Hassabis ，DeepMind创始人兼CEO

Ilya Sutskever ，OpenAI研究总监、联合创始人

Yann LeCun， Facebook AI 研究室总监，纽约大学教授

Yoshua Bengio ，蒙特利尔大学、蒙特利尔学习算法研究所所长

Stuart Russell，加利福尼亚大学伯克利分校计算机科学教授，智能系统中心主任，《人工智能：一种现代方法》的合著者

Peter Norvig ，谷歌研究总监

Ray Kurzweil ，谷歌研究总监；发明家、作家和未来学家，著有《奇点临近》

Jeff Dean， 谷歌Google Brain项目负责人，谷歌高级研究员，AAAI成员

Tom Gruber， 苹果公司；Siri的共同创始人，Siri 项目CTO和设计副总裁

Nils J. Nilsson，AAAI前任主席，斯坦福大学Emeritus Kumagai工程教授

Francesca Rossi， IBM&Padova，计算机科学教授，IJCAI主席、AAAI委员会AI与伦理问题影响委员会联席主席，ACM成员

总共有808名AI /机器人研究人员签名。

以及：

Stephen Hawking ，剑桥大学

Elon Musk，特斯拉CEO

Jaan Tallinn ，Skype 联合创始人

Sam Altman ，Y Combinator 总裁

Gill Pratt ，丰田研究院CEO

……

这23条原则分为研究问题、伦理价值观和长期问题三个部分，以确保人人都能从人工智能中受益，同时保证AI研究和发展的正确方向。

世界首个人工智能23条原则

人工智能已经为世界各地人的日常生活提供了有益的工具。人工智能技术应该在以下原则的指导下继续发展，将为未来几十年到几个世纪的人给予巨大的帮助和力量。

研究问题

1）研究目标：人工智能研究的目标应该是创造有益的智能，而不是让它像经历生物演化一样没有确定的发展方向。

2）研究资金：投资人工智能应该附带确保该研究是用于发展有益的人工智能，包括计算机科学、经济学、法律、伦理和社会研究中的棘手问题，例如：

我们如何使未来的AI系统具有高度鲁棒性，这样才能够让它们在没有故障或被黑客入侵的情况下做我们想要它们做的事情？

我们如何通过自动化实现繁荣，同时不打破资源和目的的平衡？

我们如何更新法律制度实现更大的公平和更高的效率，跟上AI的发展步伐，管控与AI相关的风险？

AI应该具有怎样的价值观，应该具有何等法律和伦理地位？

3）科学政策互联：AI研究人员和政策制定者之间应该进行有建设意义、健康的交流。

4）研究文化：应该在AI研究者和开发者中培养合作、信任和透明的文化。

5）避免竞赛（Race Avoidance）：开发AI系统的团队应积极合作，避免在安全标准方面进行削减。

伦理和价值观

6）安全：AI系统应该在整个使用周期内安全可靠，并在可行性和可用性方面有可验证的衡量标准。

7）故障透明度（Failure Transparency）：如果AI系统造成伤害，应该可以确定原因。

8）司法透明度：任何涉及司法决策的自主系统都应对其判断提供合理的解释，并由主管人权机构审核。

9）责任：先进人工智能系统的设计师和建设者是使用、滥用这些系统以及这些系统行动造成的道德影响的利益相关者，他们有责任和机会塑造这些影响。

10）价值观一致性：在设计高度自治的人工智能系统时，应该确保它们的目标和行为在整个操作过程中与人类的价值观相一致。

11）人类价值观：人工智能系统的设计和运行应与人类尊严、权利、自由和文化多样性的理念相一致。

12）个人隐私：由于AI系统能够分析和利用人类产生的数据，人类也应该有权获取、管理和控制自身产生的数据。

13）自由和隐私：AI应用个人数据，其结果不能不合理地限制人类真实或感知到的自由。

14）共享利益：人工智能技术应当惠及和增强尽可能多的人。

15）共享繁荣：人工智能创造的经济繁荣应该广泛共享，为全人类造福。

16）人类控制：应该由人类选择如何以及是否委托人工智能系统去完成人类选择的目标。

17）非颠覆（Non-subversion）：要控制先进的人工智能系统所带来的力量，应当尊重和改善社会健康发展所需的社会和公民进程，而不是颠覆这种进程。

18）AI军备竞赛：应该避免在致命的自动武器开发方面形成军备竞赛。

长期问题

19）能力注意（Capability Caution）：在没有共识的情况下，应该避免对未来AI能力上限做出较为肯定的假设。

20）重要性：先进AI可能代表了地球上生命发展历史的一大深刻变化，应该通过相应的关怀和资源对其进行规划和管理。

21）风险：人工智能系统带来的风险，特别是灾难性或有关人类存亡的风险，必须遵守与其预期影响相称的规划和缓解措施。

22）不断自我完善（Recursive Self-Improvement）：对于那些不断自我完善或通过自我复制以能够快速提高质量或数量的AI系统，必须采取严格的安全和控制措施。

23）共同利益：超级智慧只应该为广泛共享的伦理理想服务，为了全人类而不是一个国家或组织的利益而发展。

AI领袖发声

我认为，最大的危险在于，人们或许会以一种不负责任或者恶意的方式来处理AI，我指的是一些人为了个人利益会这么做。通过在全世界普及一个更加平等的社会，我认为我们可以减少这些危险。在一个充满暴力、充满不平等的社会中，AI 的误用，或者有人不负责任地使用AI的概率整体都会大增。让人人都能从AI受益是安全问题的核心。

——Yoshua Bengio, 蒙特利尔大学教授

应该设计高度自动化的AI系统，确保在操作的过程中它们的目标和行为与人类价值观保持一致。价值一致非常重要。机器人不会尝试违反人性，但是它们会对人类告诉它们要做的事进行优化。所以我们需要确保机器人所优化出的世界是我们想要的。

——Anca Dragan UC Berkeley 大学教授

致命自动化武器的军备竞赛应该被禁止。我不是战争的支持者，我认为战争是极端危险的。显然，我认为，AI 技术有着巨大潜力，但是，即便只是我们今天能掌握的技术，已经不难想象，AI可能会被用于破坏性的事务。我不希望我对这一领域的贡献，或者任何我们发展的技术会伤害到任何人，被用于开发武器或者是发动战争。哪怕我们的技术只是让武器杀伤力更强，我也不希望。

——Stefano Ermon，斯坦福大学助理教授

1980名产学研跨界人士联名

到目前为止，上述原则已由1980人签名，其中808名是 AI/机器人研究人员。签名者名单如下：

Demis Hassabis DeepMind, Founder & CEO

Ilya Sutskever OpenAI, Research Director, Co-Founder

Yann LeCun Director of AI Research at Facebook, Professor at New York University

Yoshua Bengio University of Montréal, head of the Montreal Institute for Learning Algorithms

Stuart Russell University of California, Berkeley, Professor of Computer Science, director of the Center for Intelligent Systems, co-author of the standard textbook Artificial Intelligence: a Modern Approach.

Peter Norvig Google, Research Director

Ray Kurzweil Google, Research Director; Inventor, Author and Futurist, author of The Singularity is Near and How to Create a Mind

Jeff Dean, Google, Inc., Leader of Google Brain project, Google Senior Fellow, member of AAAI, ACM, National Academy of Engineering

Tom Gruber Apple; Co-Founder, CTO and VP Design for Siri

Nils J. Nilsson, Past President of AAAI, Emeritus Kumagai Professor of Engineering, Stanford, member of AAAI

Francesca Rossi IBM & Padova, Professor of Computer Science, IJCAI President and Co-chair of AAAI committee on impact of AI and Ethical Issues, member of ACM

Bart Selman Cornell, Professor of Computer Science, co-chair of the AAAI presidential panel on long-term AI futures, member of ACM

Guru Banavar Vice President, Cognitive Computing, IBM Research

Shane Legg DeepMind, Co-Founder

Mustafa Suleyman DeepMind, Co-Founder

Leslie Kaelbling, MIT, Professor of Computer Science and Engineering, member of AAAI, ACM, IEEE RAS

Zoubin Ghahramani, University of Cambridge; Deputy Director, Leverhulme Centre for the Future of Intelligence; Cambridge Director of the Alan Turing Institute and Co-Director, Uber AI Labs, Professor, member of IEEE CS

Ramon Lopez de Mantaraa, Spanish National Research Council, Director, AI Research Institute, Barcelona, EurAI Fellow, AAAI Robert S. Engelmore award, EurAI Distinguished Service Award, member of AAAI, EurAI (European Association of Artificial Intelligence)

David Parkes, Harvard University, Professor of Computer Science, AAAI Fellow, member of ACM

Mark Mattingley-Scott, IBM, IEEE SMC, University Osnabrueck, Principal, member of IEEE CIS, IEEE RAS

Pierre Marquis, Artois University, Professor of Computer Science, EurAI Fellow

Bernhard Schölkopf, Max Planck Institute for Intelligent Systems, Director, member of ACM, IEEE CIS, JMLR, co-editor

Zlatko Lagumdzija, Professor of Computer Science, former Prime minister of Bosnia and Hercegovina and member of Club de Madrid

Peter van Beek, University of Waterloo, Professor of Computer Science, AAAI Fellow, member of AAAI

Ansgar Koene, University of Nottingham, Senior Research Fellow, member of IEEE CS

Ronald C. Arkin, Georgia Institute of Technology, Regents' Professor of Computer Science, member of AAAI, IEEE CS, IEEE RAS, IEEE Social Implications of Technology Society

Dileep George, Vicarious, Co-Founder

Grady Booch, IBM, Chief Scientist, Watson/M, member of AAAI, IEEE CS

Anca Dragan University of California at Berkeley, Assistant Professor of EECS

Roman Yampolskiy University of Louisville, Associate Professor; founder of the Cyber Security Lab

Stefano Ermon Stanford University, Assistant Professor of Computer Science

Dan Weld University of Washington, Professor of Computer Science & Engineering

Toby Walsh University of New South Wales & NICTA, Professor of AI, President of the AI Access Foundation

Tomaso Poggio Massachusetts Institute of Technology, Professor of of Brain & Cognitive Sciences, Director of Center for Brains, Minds and Machines

Steve Omohundro President of Possibility Research and Self-Aware Systems

Michael Wellman University of Michigan, Professor of Computer Science & Engineering, AAAI Fellow, member of ACM

Yutaka Matsuo University of Tokyo, Professor of Computer Science

Gregory Bonnet Ethicaa; University of Caen Normandy, Assistant Professor

Marty Tenenbaum Association for the Advancement of Artificial Intelligence; Cancer Commons

Vincent Conitzer Duke University, Professor of New Technologies

Kristinn R. Thorisson Reykjavik University, Associate Professor; Managing Director of Icelandic Institute for Intelligent Machines

Neil Lawrence University of Sheffield, Professor of Computer Science and Neuroscience; Amazon Research Cambridge

Ian Goodfellow OpenAI, AI researcher

David Poole, University of British Columbia, Professor of Computer Science, Director of Laboratory for Computational Intelligence at UBC, member of AAAI, CAIAC

Kevin Leyton-Brown, University of British Columbia, Professor of Computer Science and AI researcher, member of AAAI, ACM

Andy Chun, Hon Wai, City University of Hong Kong, Professor of Computer Science, member of AAAI

Van Hoai Tran, Faculty of computer science & engineering, Ho Chi Minh City university of technology, Associate professor of computer science

Murray Shanahan, Imperial College London, Professor of Cognitive Robotics

Rand Hindi, Snips, CEO

Tom Schaul, DeepMind, AI researcher

Peter Peer, Faculty of Computer and Information Science, University of Ljubljana, Slovenia, Professor of Computer Science (Computer Vision, Biometry), member of ACM, IEEE CS

Theresa Leigh, Singularity University, Stanford University, Harvard University, Professor of Developmental Neurogenomics; A.I. Research Scientist, member of AAAI, IEEE CIS, IEEE RAS

David L. Livingston, Virginia Military Institute, Professor of Electrical and Computer Engineering, member of IEEE CIS, IEEE RAS

Marek Rosa, GoodAI, CEO, member of AI Roadmap Institute, General AI Challenge

Adriano Barra, Università del Salento & Istituto Nazionale d'Alta Matematica, Professor ("RTD-B") of Mathematics (Neural Networks)

Ramon F. Brena, Tecnologico de Monterrey, Professor of Computer Science, member of AAAI, ACM, SMIA (Mexico)

Donald Norman, UC San Diego, Prof. Cognitive science and Director, Design Lab, member of ACM

Leonardo Garrido, Tecnológico de Monterrey, Professor of Computer Science, member of AAAI, ACM, IEEE CS, IEEE RAS, SMIA (Mexican AI Association)

Sheila McIlraith, University of Toronto, Professor of Computer Science, participant AAAI presidential panel on long-term AI futures, member of AAAI, ACM, CAIAC

Alan FT Winfield, Bristol Robotics Lab, Professor of Robot Ethics, member of IEEE RAS

Dagmar Monett, Berlin School of Economics and Law, Professor of Computer Science, member of ACM

Stephane Meystre, Medical University of South Carolina, Associate Professor and Chair, Translational Biomedical Infromatics

Catholijn M Jonker, Delft University of Technology, Professor of Interactive Intelligence, member of ACM, IEEE CS, EurAI Fellow

Alexandre Bess, Professor and CEO of AI Startup

Percy Liang, Stanford University, Assistant Professor of Computer Science

……

其他署名人士：

Stephen Hawking Director of research at Dept. of Applied Mathematics and Theoretical Physics at Cambridge, 2012 Fundamental Physics Prize laureate for his work on quantum gravity

Elon Musk Founder, CEO & CTO of SpaceX, Co-Founder & CEO of Tesla Motors, Co-Founder of OpenAI & Solar City

Jaan Tallinn Co-Founder of Skype, of the Centre for the Study of Existential Risk & Future of Life Institute

Sam Altman Co-founder of OpenAI & Loopt, president of Y Combinator

Gill Pratt CEO of Toyota Research Institute

Anthony Romero Executive director of American Civil Liberties Union

Tasha McCauley GeoSim Systems, Co-Founder of Fellow Robots

Joseph Gordon-Levitt Actor and Filmmaker

Andrew Serazin President of Templeton World Charity Foundation

Erik Brynjolfsson Massachusetts Institute of Technology, Professor of Management; Research Associate at The National Bureau of Economic Research

William MacAskill University of Oxford, Co-Founder of the Center for Effective Altruism

Huw Price University of Cambridge, Bertrand Russell Professor of Philosophy; Co-Founder of the Center for the Study of Existential Risk

Craig Calhoun President of Berggruen Institute

D. Scott Phoenix, Vicarious, Cofounder

John C. Havens, The IEEE Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomous Systems, Executive Director, member of IEEE CS

Donald Knuth, Stanford University, Professor Emeritus of The Art of Computer Programming, member of ACM, IEEE CS

Keith Abney, Cal Poly - San Luis Obispo, Senior Fellow, Ethics + Emerging Sciences Group, Philosophy Dept, member of IEEE RAS

Max Tegmark Massachusetts Institute of Technology & Center for Brains, Minds & Machines, Professor of Physics; President of Future of Life Institute

Anthony Aguirre Future of Life Institute; University of California, Santa Cruz, Professor of Physics

Irakli Beridze United Nations Interregional Crime and Justice Research Institute

Joshua Greene Harvard University, Professor of Psychology

Ryan Calo University of Washington, Assistant Professor of Law

Margaret Boden University of Sussex, Professor of Cognitive Science, member of AAAI

David Chalmers New York University, Professor of Philosophy & co-director of the Center for Mind, Brain & Consciousness

Peter Asaro The New School, Assistant Professor

Wendell Wallach Yale University

Antonio Gracias Valor Equity Partners; Tesla

Patrick Lin Stanford University, Director of the Ethics and Emerging Sciences Group

Kay Firth-Butterfield AI-Austin, Center for International Security and Law and Consortium for Law and Ethics of Artificial Intelligence and Robotics, University of Texas, Austin

Allan Dafoe Yale University, Assistant Professor of Political Science

Alexander Tamas Founding Partner of Vy Capital; Future of Humanity Institute

David Sanford LinkedIn

Juris Borzovs, University of Latvia, Professor of Information technology, member of IEEE CS

Charis Thompson, UC Berkeley and LSE, Chancellor's Professor, Science and Technology Studies

……

完整版名单，参见：https://futureoflife.org/ai-principles/ 

（澎湃新闻获新智元授权转发）