《The Age of Artificial Intelligence- Impact and Implications.pdf》由会员分享，可在线阅读，更多相关《The Age of Artificial Intelligence- Impact and Implications.pdf（36页珍藏版）》请在taowenge.com淘文阁网|工程机械CAD图纸|机械工程制图|CAD装配图下载|SolidWorks_CaTia_CAD_UG_PROE_设计图分享下载上搜索。
1、 Since AlphaGos historical victory over humans, Artificial Intelligence (AI) has attracted more and more attention from the public. In fact, AI has become a part of our everyday life and it exists almost everywhere, e.g. when you chat with Siri or read news articles on Toutiao, In a nutshell, AI is
2、an interdisciplinary science based on the integration of computer science, physiology and philosophy. To some extent, all machine behaviors augmenting human to perceive, identify, analyze and make decisions could be regarded as the applications of AI. This technology, with a history of over 60 years
3、, recently became a focus again after going through two cycles of ups and downs. As a technology company using AI to reshape the connection between humans and information, Jinri Toutiao has launched this report, The Age of Artificial Intelligence: Impact and Implications, to recognize the most influ
4、ential companies, technologies and scientists, and to understand and record the fears and expectations of AI in our society today. Preface The firms involved in comprehensive AI research are much more influential than those focused on a single field of application The open-source platforms developed
5、 by leading companies will drive the growth of applications downstream Evolving from universal chips to customized chips and then brain-like chips is the path to the future. With the further opening of open-source platforms and improvement of algorithms and chips, the key components of AI competitio
6、n will move to the competition for talents and application scenarios, from competition in technologies today. Chinese peoples AI-Confidence Index stands at 83, which shows an optimistic and rational attitude towards AI. Key Findings CONTENTS The Knowledge Map of AI The Influences of AI Firms and Tec
7、hnologies The Influences of AI Brands in Applications Scenarios AI-Confidence Index and User Profiles 01 The Knowledge Map of AI The Relations among AI-Related Fields and Their Hotness Levels The Influences of Basic AI Applications The Three Tides of AI Development The Three Key Factors for AI Devel
8、opment The Value Chain of AI Industry P1 Machine Learning Deep Learning Data Mining Database Statistics The concepts of Artificial Intelligence, Machine Learning and Deep Learning are intertwined with one another but no one is subject to another. Neural Net Algorithm and Artificial Intelligence are
9、overlapped, but the former is not simply an application of AI or its subject. Knowledge Discovery Data Souces: SAS, 2014 and PwC, 2016 Connections among the AI-Related Fields PART ONE: The Knowledge Map of AI The Family Tree of AI Pattern Recognition Artificial Intelligence Pattern Recognition Stati
10、stics Data Mining Database Deep Learning Neural Net Algorithm Machine Learning Artificial Intelligence29568611 1793671 1059994 169949 175262 20988 19180 18128 P2 PART ONE: The Knowledge Map of AI The Influences of AI-Related Fields Data Source: Toutiao Index In 1975, Paul Werbos put forward the BP a
11、lgorithm which made the learning of multi-layer artificial neural networks possible, enabling the rise of two- layer neural networks in 1986 which laid the foundation of the second tide in AI development. In 1956, The Dartmouth Artificial Intelligence Conference came up with the name and vision, and
12、 thus marked the birth of AI. 1956-1976 The First Tide of AI1976-2000 The Second Tide of AI 2000 to now The Third Tide of AI Development P3 90s The research of AI encountered the issue of insufficient funds in the 1990s. 70s Due to the bottleneck of computer performance, increasing complexity of com
13、peting, as well as data insufficiency, a lot of projects were unaccomplished in the 1970s. In 2012, Geoff Hintons team defeated Google in the ImageNet ILSVRC (Large Scale Visual Recognition Challenge), which was of great significance academically and also triggered a large-scale investment in deep l
14、earning in the industry. PART ONE: The Knowledge Map of AI Overview of AI 0 50 100 150 20042005200620072008200920102011201220132014201520162017 人工智能大数据机器学习数据科学深度学习 0 50 100 150 20042005200620072008200920102011201220132014201520162017 Artificial intelligencebig datamachine learning data sciencedeep l
15、earning Google Trends (Search with Chinese Names) Searching key words related to AI in google trends, e.g. Artificial Intelligence, Big Data, Machine Learning and Deep Learning, you can see the different focuses and how they evolve. To compare the trends of Artificial Intelligence and Big Bata, the
16、former attracted attention in both the English and Chinese contexts as early as 2004. Yet its search volume started to decline in the Chinese context around 2006-2007, with its hotness level outpaced by Big data at that time. In English language searching, the search volume of Big Data only outpaced
17、 AI around 2012. The hotness level of Deep Learning in Chinese searching was almost the same as Artificial Intelligence in 2009. However in English searching, public awareness of “Deep Learning” didnt show a significant increase until 2013. P4 Data Source: Google Trends, 2004-2017 PART ONE: The Know
18、ledge Map of AI The Rise of The Current Tide Google Trends (Search with English Names) Provides the fuel for AI Algorithm The engine of AI Big Data Reflects on the improvement of chips and cloud computing capability Computing Capability The increase of data scale, improvement of computing capability
19、 and the birth of deep learning algorithm significantly changed the AI industry and moved it forward. P5 PART ONE: The Knowledge Map of AI The Three Key Factors for AI Development Application Level Technology Level Base Level Products Industry Solution General Technology Big Data Computing Capabilit
20、y Algorithm Model Framework self-driving cars intellectual assistant robot journalists intellectual investment consultants financingtransportationretailagriculturemanufacturing voice recognition natural language processing computer vision knowledge map deep learning accumulation of underlying data c
21、hips cloud computing CNNclusteringdecision treeSVM TensorFlowCaffe2PyTorthTorchnet P6 PART ONE: The Knowledge Map of AI Structure of the AI Industry 02 AI Companies and Technologies Top 10 Most Influential AI Companies Most Influential AI Open-Source Platforms Top 10 Most Influential AI Chips The ho
22、tness level of Baidu exceeds Google in the Chinese searching. However, the hotness level doesnt translate directly into the competitiveness. Many outstanding AI companies are not listed above. 0 1000000 2000000 3000000 4000000 5000000 6000000 7000000 8000000 Examples: NVIDIA (focused on AI hardware)
23、, Jinri Toutiao (focused on intellectual information distribution), and Facebook (focused on open-source AI). The list also reveals that the companies involved in comprehensive AI research are much more influential than those focused on a single field. The top 10 most influential AI companies by Tou
24、tiao Index: Baidu, Google, Alibaba, Huawei, Apple, Tesla, Xiaomi, Jingdong, Microsoft and Didi. P7 PART TWO: AI Firms Influences Top 25 Most Influential AI Companies Data Source: Toutiao Index 531133 256707 79266 53891 2382517756 690354673893 AIs rapid development requires the advancement of open-so
25、urce platforms, to enable more middle and small-sized enterprises/developers to get involved in the research of AI. In this regard, the AI giants embarks on developing open-source platforms. In terms of the power of influence, Googles TensorFlow shows an overwhelming advantage. Meanwhile, Facebook h
26、as accelerated in this aspect these years and launched several outstanding platforms like Caffe, PyTorth and Torchnet which win wide-range recognition, though its impact is limited in the Chinese context. As the only platform launched by Chinese company, Baidus Paddle Paddle is the product developed
27、 during Andrew Ngs tenure. By developing open-source platforms, the AI giants aim to mobilize more talented scientists and engineers to participate in AIs development together. Moreover, the open-source platforms would drive up the development of the downstream of AI industry. P8 PART TWO: AI Firms
28、Influences Most Influential AI Open-Source Platforms/Frames Data Source: Toutiao Index 1827703 1650259 363808 174138 155315 80976 54187 56711 45290 18535 At present, the universal chips that are suitable for operating parallel computing such as GPU and FPGA are generally adopted to meet the new comp
29、uting demands represented by deep learning. Their importance for AI also gets confirmed in the list, and their advantages are clear: avoiding the high investment and risks in custom chips development. Yet their inborn bottlenecks in the performance and energy consumption are also distinct since it i
30、s not specialized for deep learning. As the AI application scope continues to expand, the problems would become increasingly apparent. The new computing demands in the AI age will promote the development of the new specialized computing chips: the custom chips (e.g.TPU, Cambrian and Tesla P100 GPU)
31、and the brain-like chips (e.g.TrueNorth). It could be expected that evolving from universal chips to customized chips and then brain-like chips is the path to the future. P9 PART TWO: AI Firms Influences Top 10 Most Popular Chips Data Source: Toutiao Index 334054 9933598430 87057 69503 4822046964 32
32、602 73317287 Algorithm is the core technology for AI. CNN (Convolutional Neural Network) is ranked No. 1, which is a powerful digital-image processing model making the input images into abstraction on its convolution layer. The algorithm won a good reputation for its application in the AlphaGo-Li Sh
33、ishi contest. The victory of Alpha-Go demonstrates the power and versatility of CNN. P10 Data Source: Toutiao Index PART TWO: AI Firms Influences Top 10 AI Algorithms 73% 18% 9% The AI talents largely come from Tsinghua University, Chinese Academy of Sciences, University of Science and Technology of
34、 China, Massachusetts Institute of Technology, and Peking University. It is a very well-educated group: 68% of them hold doctorate degrees. Doctorate Master Bachelor P12 Data Sources: The List of AI Talents from IYIOU.com PART TWO: AI Firms Influences The Distribution on AI Talents (Graduate School
35、and Graduate Degree) 31 1918 98 554444 141 03 AIs Influence in Applications Scenarios Top 10 Application Fields of AI Star AI Products for Consumers Star Robot Journalists Star Intellectual Assistants Star Self-driving cars / Unmanned Pilot AI as an computer technology, no matter in which scenarios
36、and fields it is applies, its nature is making technology to serve us better. There are many successful cases in some fields like finance, digital products and automobile, but the application of AI in traditional industries like agriculture, real estate and tourism still needs to be improved. A surv
37、ey by McKinsey indicates that more than 40% of companies in Chinas traditional industries have not viewed AI as their strategic priority. Only when AI gets wisely applied in these industries, rather than serving only the technology giants, its potential economic value could be maximized. P13 PART TH
38、REE: AIs Influence in Applications Scenarios Top 10 Application Fields of AI 62720950 36638065 9284495 6044750 2445465 2221795 2106120 1341860 1101020 1021410 Finance and Economics Digital Products Automobile Education Agriculture, Rural Areas and Farms Design Entertainment Real Estate Tourism Video
39、 Game Industry Data Source: Toutiao Index 0 200 400 600 800 1000 1200 1400 1600 1800 0 500000 1000000 1500000 2000000 2500000 3000000 3500000 4000000 4500000 5000000 Toutiao IndexSurvey The true value of high-end technology can only be realized when it is applied to improve the lives of people. From
40、 this perspective, simple application scenarios are exactly the best opportunities for the practices of machine learning. Though most of the recent AI products are developed for business users, those used widely by consumers have obviously gained a higher level of recognition and better reputation a
41、nd influence. For example, Googles autopilot is ranked No.1 recongnized by Toutiao Hotness Index, while Jinri Toutiao and Alipay are most recognized by the public in the questionnaire survey P14 PART THREE: AIs Influence in Applications Scenarios The Influence of AI Products Data Source: Toutiao Use
42、r Survey 32502 14230 85558542 3252 2709 2262 Robot journalist is an AI computer program used in specific fields of media reporting. As a mature programming framework has not been developed, robot journalists cannot yet cover all industries. Today robot journalists are used with a high frequency in t
43、he sports and finance channels to produce quick reports. Jinri Toutiao, Tencent and Alibaba are the leading companies in China in terms of robot- journalist technology. But Kuaibi Xiaoxin developed by the Xinhua News Agency in 2016 is the most well-known robot journalist to the public. P15 PART THRE
44、E: AIs Influence in Applications Scenarios The Influence of Robot Journalists Data Source: Toutiao Index 5063700 3111105 1627903 1346250 740161 407728 249682 Intellectual assistants, which used to exist only in science fictions, have arrived in our daily life and become well-known to the general pub
45、lic, just like Siri. By studying user behavior, such products are expected to improve the efficiency of cellphones and other similar devices. Apples Siri is the most commonly known intellectual assistant in China, followed by Baidus Xiaodu (thanks to the popular TV show Super Brain) and Microsofts X
46、iaona and Xiaoice. P16 PART THREE: AIs Influence in Applications Scenarios The Influence of Intellectual Assistant Data Source: Toutiao Index 0 10000000 20000000 30000000 40000000 50000000 60000000 70000000 80000000 90000000 It is notable that many Chinese companies appear on the Top 20 list, includ
47、ing Baidu, Didi and BYD. Autopilot (or unmanned pilot）is wisely considered as one of the most promising applications of AI, and also the most competitive one. The competitors include technology giants like Google and Apple, new automakers like Tesla and NEXTEV, traditional car manufacturers like Ben
48、z and Audi, and even NVIDIA a chip manufacturer. P17 PART THREE: AIs Influence in Applications Scenarios The Influence of the Firms in Autopilot / Unmanned Pilot Data Source: Toutiao Index 04 Public Perception of AI AI-Confidence Index The Public Perception of AI Ten Questions about AI AI-User Profiles 80.27% 37.25% 34.46% 12.72% 6.23% 5.32% NewsMoviesBooksFriends and Family Do Not Know Industry Participants News reports and movies are two main channels,accounting for 80.27% and 37.25%, respectively. Only 6.23% of respondents said they had no ideas about AI. P18 Channels t