2019年人工智能指数报告.pdf

2019 annual report ar intelligence indexRaymond Perrault (report coordinator) SRI International Yoav Shoham (chair) Stanford University Erik Brynjolfsson MIT Jack Clark OpenAI John Etchemendy Stanford University Barbara Grosz Harvard University Terah Lyons Partnership On AI James Manyika McKinsey Global Institute Juan Carlos Niebles Stanford University Project Manager and Report Editor-in-Chief Saurabh Mishra Stanford University Steering Committee Artificial Intelligence Index Report 2019 Steering CommiteeIntroduction Report Highlights Acknowledgements Chapter 1 Chapter 2 Chapter 3 Chapter 4 Chapter 5 Chapter 6 Chapter 7 Chapter 8 Chapter 9 Artificial Intelligence Index Report 2019 Table of Contents Research and Development Conferences Technical Performance The Economy Education Autonomous Systems Public Perception Societal Considerations National Strategies and Global AI Vibrancy Table of ContentsTechnical Appendix 182 4 5 9 12 37 47 71 106 127 136 146 156How to cite this Report: Raymond Perrault, Yoav Shoham, Erik Brynjolfsson, Jack Clark, John Etchemendy, Barbara Grosz, Terah Lyons, James Manyika, Saurabh Mishra, and Juan Carlos Niebles, “The AI Index 2019 Annual Report”, AI Index Steering Committee, Human-Centered AI Institute, Stanford University, Stanford, CA, December 2019. (c) 2019 by Stanford University, “The AI Index 2019 Annual Report” is made available under a Creative Commons Attribution- NoDerivatives 4.0 License (International) creativecommons/licenses/by-nd/4.0/legalcode The AI Index is as an independent initiative at Stanford Universitys Human-Centered Artificial Intelligence Institute (HAI). The AI Index was conceived within the One Hundred Year Study on AI (AI100). We thank your supporting partners We welcome feedback and new ideas for next year. Contact us at AI-Index-Reportstanford.edu. Artificial Intelligence Index Report 2019 AI Index Report Table_of_Contents 3Artificial Intelligence Index Report 2019 AI Index Report - Introduction The AI Index Report tracks, collates, distills, and visualizes data relating to artificial intelligence. Its mission is to provide unbiased, rigorously-vetted data for policymakers, researchers, executives, journalists, and the general public to develop intuitions about the complex field of AI. Expanding annually, the Report endeavors to include data on AI development from communities around the globe. Before diving into the data, it is worth noting the following about the 2019 edition of the AI Index Report: 1. This edition tracks three times as many data sets as the 2018 edition. It includes an update of previous measures, as well as numerous new ones, across all aspects of AI: technical performance, the economy, societal issues, and more. 2. This volume of data is challenging to navigate. To help, weve produced a tool that provides a high-level global perspective on the data. The Global AI Vibrancy Tool (vibrancy.aiindex) compares countries global activities, including both a cross-country perspective, as well as a country-specific drill down. Though it is tempting to provide a single ranking of countries, such comparisons are notoriously tricky. Instead, weve provided a tool for the reader to set the parameters and obtain the perspective they find most relevant when comparing countries. This tool helps dispel the common impression that AI development is largely a tussle between the US and China. Reality is much more nuanced. Our data shows that local centers of AI excellence are emerging across the globe. For example, Finland excels in AI education, India demonstrates great AI skill penetration, Singapore has well-organized government support for AI, and Israel shows a lot of private investment in AI startups per capita. 3. We are also releasing the AI Index arXiv Monitor (arxiv.aiindex), a tool to support research on current technological progress in AI via full-text searches of papers published on the pre-print repository. Given that measurement and evaluation in complex domains remain fraught with subtleties, the AI Index has worked hard to avoid bias and seek input from many communities. As part of this effort, on October 30, 2019, the Stanford HAI-AI Index Workshop: Measurement in AI Policy: Opportunities and Challenges ( hai.stanford.edu/ai-index/workshops) convened over 150 industry and academic experts from a variety of disciplines related to AI to discuss the many pressing issues that arise from data measurement of AI. The Workshop Proceedings will be available shortly here. 4 Introduction to the AI Index 2019 Report Table_of_ContentsArtificial Intelligence Index Report 2019 AI Index Report - Highlights Each of the nine chapters presents well-vetted data on important dimensions related to the activity in, and technical progress of artificial intelligence. Here is a sample of the findings. 1. Research and Development Between 1998 and 2018, the volume of peer-reviewed AI papers has grown by more than 300%, accounting for 3% of peer-reviewed journal publications and 9% of published conference papers. China now publishes as many AI journal and conference papers per year as Europe, having passed the US in 2006. The Field-Weighted Citation Impact of US publications is still about 50% higher than Chinas. Singapore, Switzerland, Australia, Israel, Netherlands, and Luxembourg have relatively high numbers of Deep Learning papers published on arXiv in per capita terms. Over 32% of world AI journal citations are attributed to East Asia. Over 40% of world AI conference paper citations are attributed to North America. North America accounts for over 60% of global AI patent citation activity between 2014-18. Many Western European countries, especially the Netherlands and Denmark, as well as Argentina, Canada, and Iran show relatively high presence of women in AI research. 2. Conferences Attendance at AI conferences continues to increase significantly. In 2019, the largest, NeurIPS, expects 13,500 attendees, up 41% over 2018 and over 800% relative to 2012. Even conferences such as AAAI and CVPR are seeing annual attendance growth around 30%. The WiML workshop has eight times more participants than it had in 2014 and AI4ALL has 20 times more alumni than it had in 2015. These increases reflect a continued effort to include women and underrepresented groups in the AI field. 3. Technical Performance In a year and a half, the time required to train a large image classification system on cloud infrastructure has fallen from about three hours in October 2017 to about 88 seconds in July, 2019. During the same period, the cost to train such a system has fallen similarly. Progress on some broad sets of natural-language processing classification tasks, as captured in the SuperGLUE and SQuAD2.0 benchmarks, has been remarkably rapid; performance is still lower on some NLP tasks requiring reasoning, such as the AI2 Reasoning Challenge, or human-level concept learning task, such as the Omniglot Challenge. Prior to 2012, AI results closely tracked Moores Law, with compute doubling every two years. Post-2012, compute has been doubling every 3.4 months. 4. Economy Singapore, Brazil, Australia, Canada and India experienced the fastest growth in AI hiring from 2015 to 2019. 5 AI Index 2019 Report Highlights Table_of_Contents Artificial Intelligence Index Report 2019 AI Index Report - Highlights In the US, the share of jobs in AI-related topics increased from 0.26% of total jobs posted in 2010 to 1.32% in October 2019, with the highest share in Machine Learning (0.51% of total jobs). AI labor demand is growing especially in high-tech services and the manufacturing sector. The state of Washington has the highest relative AI labor demand. Almost 1.4% of total jobs posted are AI jobs. California has 1.3%, Massachusetts 1.3%, New York 1.2%, the District of Columbia (DC) 1.1%, and Virginia has 1% online jobs posted in AI. In the US, the share of AI jobs grew from 0.3% in 2012 to 0.8% of total jobs posted in 2019. AI labor demand is growing especially in high-tech services and the manufacturing sector. In 2019, global private AI investment was over $70B, with AI-related startup investments over $37B, M AI4All board; Bloomberg Philanthropies; WiML board; Pedro Avelar (UFRGS), Dhruv Batra (Georgia Tech / FAIR); Zoe Bauer; Sam Bowman (NYU); Cody Coleman (Stanford); Casey Fiesler (University of Colorado Boulder); Brenden Lake (NYU); Calvin LeGassick; Natalie Garrett (University of Colorado Boulder); Bernard Ghanem (King Abdullah University of Science and Technology); Carol Hamilton (AAAI); Arthur Jago (University of Washington Tacoma); Zhao Jin (University of Rochester); Lars Kotthoff (University of Wyoming); Luis Lamb (Federal University of Rio Grande do Sul); Fanghzhen Lin (Hong Kong University of Science and Technology); Don Moore (UC Berkeley Haas School of Business), Avneesh Saluja (Netflix), Marcelo Prates (UFRGS), Michael Gofman (University of Rochester); Roger McCarthy (McCarthy Engineering); Devi Parikh (Georgia Tech / FAIR); Lynne Parker (White House Office of Science and Technology Policy); Daniel Rock (MIT); Ayush Shrivastava (Georgia Tech College of Computing); Cees Snoek (University of Amsterdam); Fabro Steibel (ITS-RIO); Prasanna Tambe (Wharton); Susan Woodward (Sandhill Econometrics); Chenggang Xu (Cheung Kong Graduate School of Business). Matthew Kenney (Duke University) built the arXiv search engine tool. Tamara Pristc (Stanford) and Agata Foryciarz (Stanford) provided research contributions. Biswas Shrestha (Stanford) supported editing. Special thank you to Monique Tuin (McKinsey Global Institute) for invaluable comments and feedback. Michael Chang (graphic design and cover art), Kevin Litman-Navarro (data visualization), Ruth Starkman (editor) and Biswas Shrestha (Stanford) were integral to the daily production of the report.