Ce diaporama a bien été signalé.
Le téléchargement de votre SlideShare est en cours. ×

Panel: AI for Social Good - Fairness, Ethics, Accountability, and Transparency

Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité
Publicité

Consultez-les par la suite

1 sur 51 Publicité

Panel: AI for Social Good - Fairness, Ethics, Accountability, and Transparency

As we begin to harness the power of artificial intelligence, machine learning, and data science in our everyday lives, we also raise complex ethical and social questions associated with bias, fairness, and transparency of algorithmic intelligence. In this panel we get into the thick of the issue. How can we best use AI with shared responsibilities between humans and systems? How can we balance the need for efficiency and exploration with fairness and sensitivity to users? How do we ensure that individuals and communities can trust these systems? Join our discussion to enrich your understanding of human-AI interaction, and how these questions will be answered in AI research, education and policies, as we strive to improve the human condition.

As we begin to harness the power of artificial intelligence, machine learning, and data science in our everyday lives, we also raise complex ethical and social questions associated with bias, fairness, and transparency of algorithmic intelligence. In this panel we get into the thick of the issue. How can we best use AI with shared responsibilities between humans and systems? How can we balance the need for efficiency and exploration with fairness and sensitivity to users? How do we ensure that individuals and communities can trust these systems? Join our discussion to enrich your understanding of human-AI interaction, and how these questions will be answered in AI research, education and policies, as we strive to improve the human condition.

Publicité
Publicité

Plus De Contenu Connexe

Diaporamas pour vous (20)

Similaire à Panel: AI for Social Good - Fairness, Ethics, Accountability, and Transparency (20)

Publicité

Plus par Amazon Web Services (20)

Panel: AI for Social Good - Fairness, Ethics, Accountability, and Transparency

  1. 1. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T AI for Social Good - Fairness, Ethics, Accountability, and Transparency
  2. 2. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Opinions expressed are solely our own and do not express the views or opinions of Amazon or AWS
  3. 3. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Machine Learning 1980s A technique that allows computer to perform task without being explicitly programmed AI, Machine Learning, and Deep Learning Artificial Intelligence 1950s Any techniques that allows computer to mimic human intelligence Turing Test Perceptron Deep Learning 2010s A subfield of machine learning that uses neural network to learning
  4. 4. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T 3 Types of Machine Learning Supervised Machine Learning Task driven • Training Data: (X,Y) (Features, Labels) • Predict: Y, minimizing some loss • Classification, Regression Unsupervised Machine Learning Data driven • Training Data: X (features only) • Find similar points in high-dim X-space • Clustering Reinforcement Learning Decision making • Training data: (State, Action, Reward) • Maximize long term rewards • Robotics, games
  5. 5. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T
  6. 6. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T AI for Social Good (Collaboration: AWS, NSF and University of Nevada) https://www.unr.edu/nevada-today/news/2019/big-data-wildfires
  7. 7. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Source: Weather Channel Machine Learning for Improving Disaster Management and Response Session ID: 301069 - Artificial Intelligence and Machine Learning in Research
  8. 8. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T AI for Social Good
  9. 9. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Evolving AI Capabilities
  10. 10. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Survey of Judges: Case for Human Decision making William Austin and Thomas A. Williams III, ‘A survey of judges’ responses to simulated legal cases: research note on sentencing disparity’, Journal of Criminal Law and Criminology , vol. 68, no. 2, 1977, pp. 306–310. Algorithmic Learning:
  11. 11. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Applicability in various areas:
  12. 12. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T
  13. 13. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Fairness • Protected Attributes : like race, gender, age, religion and their proxies • Classification parity, meaning that common measures of predictive performance (e.g., false positive and false negative rates) • Outcomes are independent of protected attributes. Transparency Accountability Explainability Example: DARPA’s program on explainable AI. https://www.darpa.mil/program/explainable-artificial-intelligence
  14. 14. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Prem Natarajan is a Vice President in Alexa and leads a multidisciplinary science, engineering, and product organization which improves customer experience worldwide through advances in natural language understanding, entity linking and resolution, and related machine learning technologies. Before joining Amazon, he was senior vice dean of engineering at the University of Southern California where he led nationally influential DARPA and IARPA sponsored research efforts in biometrics/face recognition, OCR, natural language processing, media forensics, and forecasting. Prior to that, he served as executive vice president and principal scientist for speech, language, and multimedia at Raytheon BBN Technologies.
  15. 15. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T AI for Social Good
  16. 16. AMAZON.COM INC. 2019. ALL RIGHTS RESERVED. AI for Social Good  Social “good” comes in many forms  Better education  Faster, cheaper drug discovery  More effective policy making  Predicting and responding to natural disasters, epidemics  And much more ….  Commonly accepted attributes of “AI for Social Good”  Collaboration of multiple disciplines, especially social sciences and AI  Public-private partnership + nonprofits-academia-industry collaboration  Open access to technological resources  Considerations of bias, fairness, and accountability of ML algorithms
  17. 17. AMAZON.COM INC. 2019. ALL RIGHTS RESERVED. Current State  Growing investments across academia, industry, and government  Several academia-based centers of “AI for Social Good” or “AI in Society” have emerged in recent years with diverse themes ranging from algorithms to policy  Industry initiatives span in-house efforts and extramural community creation efforts such as the NSF-Amazon Fairness Program for funding fairness research in academia  Substantial Government investments – e.g. DARPA LORELEI, Memex, World Modelers, XAI and many other programs  Emergence of conferences and workshops  FATML – Fairness, Accountability and Transparency in ML  AI for Social Good workshop at NeurIPS
  18. 18. AMAZON.COM INC. 2019. ALL RIGHTS RESERVED. Current State – AI Stack View Apps Toolkits ML Dev Environments Algorithms Compute / Storage Mostly Open source (e.g. MXNet) but includes dev environments like Alexa Skills Kit Requires targeted funding for fairness, transparency, etc. Requires Funding Cost-effective models of access
  19. 19. AMAZON.COM INC. 2019. ALL RIGHTS RESERVED. Alexa Skills for Social Good  Organized contest in 2018 to encourage creation of Alexa skills for social good  Red Cross skills: hurricane alerts, scheduling blood donations, and first aid  Environmental consciousness skills: recycle Game, EVIE assistant, compost tracking, bike sharing  Access skills: My Talking Newspaper, Safe and Well (check on status of relatives)  Language Preservation (with the Alexa Cleo Skill)  Cleo skill harnesses the expertise of multilingual Alexa users to teach Alexa new languages or dialects. Through a crowdsourcing model, users can help expand Alexa to new locales and languages, bringing the technology to more people around the world.  Users have taught Alexa languages such as Hindi, Korean, Russian, Klingon and many more.  We are conducting an internal pilot to evaluate programs to support language preservation with Indigenous languages such as Lakota and Ojibwe.
  20. 20. AMAZON.COM INC. 2019. ALL RIGHTS RESERVED. Doing Well by Doing Good* “How People with Disabilities Are Using AI to Improve Their Lives” “It was the first time since he was a toddler playing with a rattler that he was able to interact with something all by himself,” James says. “This Echo device goes way beyond ordering groceries or looking up a recipe for us." --- NPR Nova 30 January 2019 “How the Alexa Robot brought internet-based learning to a remote village school in Maharashtra” “….. people on ground zero have emerged as change-makers themselves with a little help from Amazon devices. Here’s one such story that is nothing but a triumph of human imagination.” “In the hot, dry, and dusty village of Warud in Maharashtra’s Amravati district, a 31-year-old schoolteacher is using Alexa to impart lessons to kids of farmers and labourers employed in the vicinity.” --- Yourstory.com and The Hindu newspaper, 4 Feb 2019 *Prof. Andrew Lo at re:MARS 2019
  21. 21. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Jim Kurose Assistant Director, CISE, National Science Foundation Dr. Jim Kurose is an Assistant Director of the National Science Foundation, where he leads the Directorate for Computer and Information Science and Engineering (CISE). With an annual budget of nearly $1B, CISE’s mission is to uphold the nation's leadership in scientific discovery and engineering innovation through its support of fundamental research in computer and information science and engineering, transformative advances in cyberinfrastructure, and preparation of a diverse computing-capable workforce. Jim also co-chairs the Networking and Information Technology Research and Development (NITRD) Program, the Subcommittee on Machine Learning and AI, and the Subcommittee on Open Science of the National Science and Technology Council (NSTC), facilitating the coordination of these research and development efforts across Federal agencies. Recently, Jim also served as the Assistant Director for Artificial Intelligence in the US Office of Science and Technology Policy (OSTP). Jim is on leave from the University of Massachusetts, Amherst, where he is Distinguished University Professor of Computer Science
  22. 22. Information & Intelligent Systems Computing & Communication Foundations Computer & Network Systems Advanced Cyberinfrastructure Panel: AI for Social Good - Fairness, Ethics, Accountability, and Transparency AWS Public Sector Summit Jim Kurose Assistant Director, NSF Computer & Information Science & Engineering Federal AI R&D Activities: a view from NSF
  23. 23. AI: ongoing US government activities AI Executive Order (Feb 2019) HSST AI Roundtable (May 2019) Congress Senate, House legislative activities AI Convening @ NSF (May 2019)  Envisioning National AI R&D Institutes  Policy and principles  Objectives  Roles and responsibilities  Federal Investment in AI R&D  Data, Computing for AI R&D  Guidance for Regulation of AI Applications  AI and the American workforce  Action Plan for Protection of the United States Advantage in AI
  24. 24. AI principles Principles for responsible stewardship of trustworthy AI  Inclusive growth, sustainable development and well-being  Human-centred values and fairness  Transparency and explainability  Robustness, security and safety  Accountability National policies and international co-operation for trustworthy  Investing in AI R&D  Fostering a digital ecosystem for AI  Building human capacity, preparing for labour market transformation  International cooperation for trustworthy AI OECD Principles on AI, May 22, 2019
  25. 25. Fairness in the AI System Lifecycle Artificial Intelligence in Society, June 12, 2019
  26. 26. NSF Leadership in AI NSF invested nearly $450M in AI research (core, applications, systems, infrastructure) in FY18 $ Thought Leadership Across USG Innovative Programmatics NSTC Select Committee on AI NSTC Subcommittee on ML & AI NSTC AI Interagency Working Group (under NITRD): 2016, 2019 AI R&D Strategic Plans OSTP Assistant Director(s) for AI International: OECD, G7 Research Funding
  27. 27. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Patricia Flatley Brennan, RN, PhD Director, National Library of Medicine National Institutes of Health Prem Natarajan is a Vice President in Alexa and leads a multidisciplinary science, engineering, and product organization which improves customer experience worldwide through advances in natural language understanding, entity linking and resolution, and related machine learning technologies. Before joining Amazon, he was senior vice dean of engineering at the University of Southern California where he led nationally influential DARPA and IARPA sponsored research efforts in biometrics/face recognition, OCR, natural language processing, media forensics, and forecasting. Prior to that, he served as executive vice president and principal scientist for speech, language, and multimedia at Raytheon BBN Technologies.
  28. 28. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T The National Institutes of Health NEI NCI NHLBI NLM NINDS NIMH NIAMS NINR NCCIH NHGRI NIA NIAAA NIAID NICHD NIDCD NIDCR NIDDK NIDA NIEHS OD NIBIB NIMHD NCATS CIT CC CSR FICNIGMS
  29. 29. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T BIOMEDICAL CHALLENGES -Cardiovascular Health - Gene Therapy - Alzheimer’s Disease - Lifespan Development - BRAIN
  30. 30. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T and many others Open Science Open Data at NIH STRIDE S
  31. 31. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T 31 NIH’s Longstanding Commitment to Data & Resource Sharing 20042003 2007 20142008 NIH Model Organism Policy NIH Genome-wide Association (GWAS) Policy 2012 NIH Public Access Policy (Publications) Big Data to Knowledge (BD2K) Initiative NIH Genomic Data Sharing (GDS) Policy White House Initiative (“OSTP Memo”) Increasing Access Results Fed- Funded Sci Research 2015 2017 NIH Public Access Plan NIH Data Sharing Policy Modernization of NIH Clinical Trials Request for Information (RFI) on Data Sharing NIH Data Commons Pilot 2016 Cancer Moonshot 2013 NLM Strategic Plan NIH Data Science Strategic Plan 2018 RFI on Policy Provisions for Data Management and Sharing HHS Rule and NIH Policy on Clinical Trial Results Dissemination NIH New Models for Data Stewardship and STRIDES Initiative 21st Century Cures Act NIH All of Us Research Program Precision Medicine Initiative Adapted from NIH Office of Science Policy
  32. 32. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T NLM as a platform for biomedical discovery & data-powered health New ways to reach users with new information and new tools A workforce prepared to advance data-driven discovery & data-powered health NLM Transforming Information into Discovery
  33. 33. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Fostering a sphere of discovery: digital research objects Protocols Funding Code Models Clinical Data Literature Study Data People Pathways Instruments
  34. 34. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Critical Issues in the Development and Use of AI for Biomedical Discovery Rigor and Reproducibility Protection of participant privacy Discoverability of data sets Data Stewardship: Preservation, Sustainability, Sharing etc.
  35. 35. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Tomas Diaz de la Rubia Vice President for Discovery Park, Purdue University Tomás Díaz de la Rubia is Purdue University's vice president for Discovery Park. In this position, his responsibilities include building upon Discovery Park's foundation of excellence, which has enabled high-impact research that crosses traditional academic boundaries. Prior to Purdue, Tomás served as chief research officer and deputy laboratory director for science and technology at the Lawrence Livermore National Laboratory (LLNL) in California, where he was responsible for the science and technology foundations of the laboratory’s $1.6 billion research program. In this capacity, he oversaw a $300M program of basic and applied research, and was responsible for the Laboratory’s industrial partnerships and technology commercialization. Tomás has published more than 150 peer-reviewed articles and has co-edited several books and conference proceedings. He is a fellow of the American Physical Society and of the American Association for the Advancement of Science and served as an elected member of the board of directors of the Materials Research Society, and vice-chair of the division of computational physics of the American Physical Society. He holds a bachelor's degree and a doctorate in physics from The State University of New York, Albany.
  36. 36. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Azer Bestavros Director, Hariri Institute for Computing, Boston University Azer Bestavros is Warren Distinguished Professor of CS and Founding Director of the Hariri Institute for Computing at BU, an incubator for high-risk, high-reward cross- disciplinary projects. His current research is on the design and implementation of scalable secure multiparty computation platforms to enable analytics over private data. Funded by over $30M from government and industry, his research yielded 18 PhD theses, 8 patents, 2 startups, and hundreds of papers with over 20,000 citations.
  37. 37. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Sharing Knowledge without Sharing Data Towards Fairness, Ethics, Accountability, and Transparency
  38. 38. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T ML: Actionable Knowledge from Data 𝑓 𝑥1, 𝑥2, 𝑥3, …
  39. 39. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T ML: The Bad News…  Repeated queries to model leak inputs.  Adversaries can pollute input to reveal data.  Adversaries can steal the learned models.
  40. 40. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Differential Privacy: The Promise Credit: https://tinyurl.com/y6x9kdyx
  41. 41. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T ML & DP: The Good News CloakwithDP
  42. 42. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T ML: Actionable Knowledge from Data g 𝑥1, 𝑥2, 𝑥3, …
  43. 43. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Multi-Party Computation: The Promise K = 𝑓 𝑥1, 𝑥2, 𝑥3, …
  44. 44. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Dec 11 2013 GOAL 3: Evaluating Success Employers agree to … contribute data to a report compiled by a third party ... Employer-level data would not be identified in the report.
  45. 45. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T
  46. 46. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T November, 2017 ++
  47. 47. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T The congresswoman, who had signed onto a bill addressing income disparity between men and women, was impressed by the relevance he outlined. “It’s linking it back for the members of Congress,” Clark said. “Nobody would think, oh, the Paycheck Fairness Act, how is that tied into NSF funding?” 2014  2018 2017 2015 “[MPC] has never been used for public good. Here, we’re beginning to show how to use this sophisticated computer science research for public programs.” BWWC co-chair Evelyn Murphy 2014 “This [is] the first time actual wage data has been reported both anonymously and voluntarily. This is a groundbreaking moment in tackling the gender gap.” Mayor Marty Walsh
  48. 48. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T
  49. 49. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T NSF Program on Fairness in Artificial Intelligence in Collaboration with Amazon (FAI) https://www.nsf.gov/pubs/2019/nsf19571/nsf19571.htm
  50. 50. Thank you! © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T Sanjay Padhi: sanpadhi@amazon.com
  51. 51. © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T © 2019, Amazon Web Services, Inc. or its affiliates. All rights reserved.P U B L I C S E C TO R S U M M I T

×