WIPO-2019年科技趋势之人工智能报告（英文）-2019.6-158页.pdf

WIPOTechnology Trends 2019ArtificialIntelligenceWIPOTechnology Trends 2019ArtificialIntelligenceThe user is allowed to reproduce,distribute,adapt,translate and publicly perform this publication,including for commer-cial purposes,without explicit permission,provided that the content is accompanied by an acknowledgement that WIPO is the source and that it is clearly indicated if changes were made to the original content.Suggested citation:WIPO(2019).WIPO Technology Trends 2019:Artificial Intelligence.Geneva:World Intellectual Property Organization.Adaptation/translation/derivatives should not carry any official emblem or logo,unless they have been approved and validated by WIPO.Please contact us via the WIPO website to obtain permission.For any derivative work,please include the following disclaimer:“The Secretariat of WIPO assumes no liability or responsibility with regard to the transformation or translation of the original content.”When content published by WIPO,such as images,graphics,trademarks or logos,is attributed to a third-party,the user of such content is solely responsible for clearing the rights with the right holder(s).To view a copy of this license,please visit https:/creative-commons.org/licenses/by/3.0/igo/The designations employed and the presentation of material throughout this publication do not imply the expression of any opinion whatsoever on the part of WIPO concerning the legal status of any country,territory or area or of its authorities,or concerning the delimitation of its frontiers or boundaries.This publication is not intended to reflect the views of the Member States or the WIPO Secretariat.The mention of specific companies or products of manu-facturers does not imply that they are endorsed or recom-mended by WIPO in preference to others of a similar nature that are not mentioned.Photo credit:cover montage created with images by Margarita Lyr/iStock/Getty Images Plus and Daria Dombrovskaya/iStock/Getty Images PlusPrinted in Switzerland WIPO,2019First published 2019World Intellectual Property Organization 34,chemin des Colombettes,P.O.Box 18 CH-1211 Geneva 20,SwitzerlandISBN:978-92-805-3007-0Attribution 3.0 IGO(CC BY 3.0 IGO)Artificial intelligence is a new digital frontier that will have a profound impact on the world,transforming the way we live and work.WIPO Director General,Francis GurryPreface7Foreword8About the contributors10Acknowl-edgments12Executive summary131 Introduction The past,present and future of AI:what research and innovation trends can reveal;the data used in this report and how it is analyzed;and a scheme for categorizing AI technologies.182 Trends in artificial intelligence Overall trends emerging from the data and analysis of changes over time,by region and industry and the most prominent entities.303 Evolution of AI patent applications and scientific publications The historical development of AI innovation:analysis of trends in patents and scientific literature since the emergence of AI,and breakdown by techniques,functional applications and application fields.384 Key players in AI patentingThe top applicants for AI patents:how companies and universities/public research organizations compare,which entities are most active in each area and where they are filing.585 WIPO Technology Trends 20195 Geography of patent filingsComprehensive analysis of geographical trends,based on the offices where patents are filed,including both first and subsequent filings.826 Market trends related to AI Analysis of the use of AI technologies:data on acquisitions,funding,open source and patent litigation and oppositions.1047 Key issues arising from AI and policy responsesViews from AI experts on the key policy and regulatory issues raised by AI applications,plus examples of policies proposed or enacted in various jurisdictions.1208 The future of AI and the IP systemThe opportunities presented by AI for business,society and culture,and how AI and IP rights interact with each other in the promotion of innovation.138Selected AI categories and terms146Furtherreading150This new report aims to shed light on the trends in innovation in AI since the field first developed in the 1950s.WIPO Director General,Francis Gurry7 WIPO Technology Trends 2019PrefaceI am pleased to present the first report in a new flagship series,WIPO Technology Trends.This first edition features artificial intelligence(AI)as the theme.It is a fitting topic for the launch of the series as AI is a cutting-edge technology with impacts on a wide range of businesses and activities.AI is fast becoming part of our everyday lives,changing how we work,shop,travel and interact with each other.Yet we are only at the beginning of discovering the many ways in which AI will have an impact on and indeed challenge business,society and culture.There are numerous misconceptions and misgivings about the nature of AI,and in particular the challenge it poses to humankind.Given these widely held reservations and concerns,it is essential to have a factual basis for policy discussions about innovation in AI.Through this report,we hope to contribute evidence and bring clarity to this important area of debate.Based on a detailed study of patents and other information,this new report aims to shed light on the trends in innovation in AI since the field first developed in the 1950s.Research has involved the analysis of patent data related to AI inventions,as well as data on scientific publications,litigation filings and acquisition activity.These data findings are discussed in detail and accompanied by commentary and industry perspectives from more than 20 of the worlds leading experts in AI;more complete datasets,patent search methodology and expert contributions are openly available on the WIPO website.We hope that this report will prove an invaluable resource for businesses,researchers and policymakers in the field,as well as for those general readers who want to find out how AI research has so far developed and where it is going.The analysis offers unique insights into trends in AI techniques(i.e.,the different approaches used in AI,such as machine learning and fuzzy logic),AI functional applications(such as natural language processing,speech processing and computer vision)and AI application fields(i.e.,those industries and other sectors in which AI innovation is being put into practice).One of the most striking findings of the report is that 50 percent of all AI patents have been published in just the last five years a remarkable illustration of how rapidly innovation is advancing in this field.This report identifies the key players in AI from both the corporate and public sectors across different research areas and industries.Furthermore,the analysis of the data and the inputs from AI experts address many of the policy issues raised by AI,such as the regulation and control of data,the incentivization of further research,the role of intellectual property(IP)protection and the development of human-centered and ethical AI to benefit all.Together,the analysis of technological trends presented in this report and the voices of AI experts collected here are a valuable new addition to a growing knowledge base on AI.I hope that this contribution will help to shift debate away from speculative interpretation and toward evidence-based projections,thereby informing global policymaking on the future of AI,its governance and the IP framework that supports it.Francis GURRY Director General8 ForewordForewordAndrew Ng CEO,Landing AI and deeplearning.ai I remember watching the final chess match between Deep Blue and Gary Kasparov while I was a graduate student at MIT in the 1990s.When Kasparov resigned,I jumped out of my chair in excitement the AI community had finally triumphed over the human chess champion!This moment stands out for me,among many milestones in the development of AI,such as the rise of deep learning,lessons in scalability and the DARPA Grand Challenge.AI is automation on steroids,and we had“automated”the process of playing chess.I was once captain of my high school chess team and used to play competitively,but at that moment I was happy to retire and leave it to the computers.This was not a sad event for me;on the contrary,I was thrilled to give up chess this way.As AI continues to develop and transform industry after industry,I hope it will create many more similarly joyful experiences.We have a duty to address serious issues relating to unemployment and inequality,but we should also never lose sight of the value and possibility that AI brings.This report illustrates some of those possibilities and where they are arising.One of its findings is that deep learning is the biggest and fastest growing technique in AI.When I was at Stanford in the 2000s,my PhD student,Adam Coates,came into my office with a chart showing that the more data you fed to a neural network,the better the neural network performed.We started looking for the biggest computers we could find.That discovery,and the idea of using distributed computing to scale up neural networks,led to the creation of Google Brain in 2011.Since then,deep learning has made great progress as weve collected more data and created the powerful computers to make it work.But we still have a long way to go in the field.For example,a toddler can usually recognize a cat after just one encounter,but a computer still needs more than one example to learn.We need to find ways to train computers on training datasets as small as 100,or even 10.Manufacturing lines,for example,use computer vision to check for defects in parts but hopefully they will never have a million examples of defects to train on.Effective“unsupervised learning”learning without labelled data remains a holy grail of AI.Even without this“holy grail,”AI is already creating massive economic value in the world today.In covering AI,the media tends to focus on images,speech and natural language processing because those types of data are very human.Everyone can understand what it means for a computer to identify a cat or recognize your speech.But a lot of the economic value today is driven through deep learning on“structured data.”Think of this as machine learning applied to massive Excel spreadsheets(only bigger than the biggest Excel spreadsheet in the world).This data for example,what trucks should be dispatched when,what products to recommend to what user is more specific to individual companies and industries.Structured data gets less attention because its not as visual or intuitive.But it is driving massive value today and will continue as incumbent companies across all industries transform themselves with AI.With this potential to transform every industry and create so much economic value,AI presents just as great a technological shift as the Internet did.In the Internet era,companies including Microsoft and Apple saw the Internet trend early,and their leaders made strategic decisions to become true Internet companies.9 WIPO Technology Trends 2019This meant investing in the right technology,building teams,developing a digital strategy and adapting how they communicated about their products.We are seeing a small number of incumbent companies undergo a similar transformation in the AI era.Google and Baidu were leaders in hiring AI talent and building pilot AI projects;they are now fully AI-driven companies.But AI transformation is not limited solely to Internet search companies.The biggest untapped opportunities today lie outside of the software industry in industries such as agriculture,healthcare and manufacturing.Every organization can leverage its data and domain expertise to build pilot AI projects.They will then be able to build internal AI teams that push their industries forward and create unique AI solutions.I recently released the AI Transformation Playbook to help organizations in all industries navigate their own AI transformations.In order to help more people and businesses harness the potential of AI,we should consider the following three actions:1.Build more publicprivate partnerships:When companies and universities or government organizations work together,both sides can benefit.For example,university researchers can access more data and understand relevant problems,and corporations can understand the latest breakthroughs in technology.2.Continue promoting the free and open sharing of AI knowledge and resources:Companies have joined universities as leading forces for publishing free and open AI research.The arXiv repository has made a huge difference to the AI revolution,as has the hosting service GitHub because it dramatically accelerates the free dissemination of ideas.Other fields can learn this lesson and move away from paywalled journals.3.Promote increased understanding of AI:Today,you can get a PhD to learn AI,but its not the only option.Traditional degrees,jobs that allow you to learn,and at-home online learning programs(such as MOOCs)are all good ways for people to learn and work on important problems.The more widely accessible information we have,the more flexibility people will have to learn.Today,a significant amount of AI research and education is taking place in the United States and China.These countries are home to many of the best universities in the world,and their governments have provided funding and created thoughtful regulations that enable innovation.But the United States and China have also built incredible business ecosystems.Its very difficult for other countries,even those with great education,to compete with the business,engineering and investing talent of those two countries.To close this gap,and work towards decentralizing the concentration of AI technology,we must invest in education.Governments should invest heavily in educating their citizens and enter publicprivate partner-ships to adopt AI-powered systems safely.AI will transform every facet of society.It brings tremendous promise to improve our lives and the world we live in,but it will require the creation of an AI ecosystem to ensure long-term,sustainable growth.As this report highlights,AI technology and innovation has until now been focused on a small number of regions and organizations.In building a fairer and more equitable AI-driven society,we must empower businesses,governments and citizens who may be impacted by automation to ensure that the benefits of AI are widely shared.10 About the contributorsSeth G Benzell is postdoctoral associate at the Massachusetts Institute of Technology(MIT)Initiative on the Digital EconomyNick Bostrom is Director of the Future of Humanity Institute and author of Superintelligence:Paths,Dangers,StrategiesErik Brynjolfsson is Director at the MIT Initiative on the Digital EconomyYoon Chae is a senior associate Baker McKenzieFrank Chen is a partner at Andreessen HorowitzMyriam Ct is Director of AI for Humanity at Montreal Institute for Learning Algorithms(Mila)Boi Faltings is Professor of AI and Director o