Artificial Intelligence - Holland Innovation Network Special - RVO

Page created by Stephanie Mullins
 
CONTINUE READING
Artificial Intelligence - Holland Innovation Network Special - RVO
Holland Innovation Network Special

Artificial
Intelligence
Artificial Intelligence - Holland Innovation Network Special - RVO
Table of contents
                                                        AI can be the game changer and

  20
                                                        everybody wants to be first

                                                         United States

  China                                                                             8
China has the potential to become the world’s largest
supplier of Artificial Intelligence products
                                                           United States
                                                           Silicon Valley

                                                          16
  Germany                               28              Silicon Valley accounted for 40% of
                                                        the world’s total investments

Cooperation across sectors is key in Germany

                                                                                         AI special
Artificial Intelligence - Holland Innovation Network Special - RVO
1

                           3                    8                      12                        16
The Netherlands                United States        United States - Boston   United States - Silicon Valley

                          20                   28                     32                        35
China                          Germany              France                   United Kingdom

                          38                   40                    42                         48
South Korea                    Brazil               Taiwan                   India

                          52                   56                    60                         62
Japan                          Russia               Turkey                   Singapore

Artificial Intelligence
Artificial Intelligence - Holland Innovation Network Special - RVO
2                                                                                                                                                            Preface

                                  Dear reader,
                                Artificial intelligence is or will be a game-changer in many or all sectors of human life, just as
                                the Industrial Revolution and the advent of ICT have been.

                                The ultimate scope of AI is as yet unknown but seems infinite.

                                With pleasure we present this special issue magazine of the Holland Innovation Network on
                                the international trends and developments in the field of Artificial Intelligence in those
                                countries where the Holland Innovation Network is based.
                                The Innovation Attachés give an overview of the scientific and technological developments,
                                knowledge institutes and the government policy in their countries.

                                The Holland Innovation Network is there for you: companies, knowledge institutes en
                                governments with ambitions in the field of international innovation and cooperation.
                                We have offices in the following countries: France, Germany (and Switzerland), Israel, Sweden,
                                UK, Russia, USA (including Canada), Brazil, China, India, Japan, Taiwan, Singapore and
                                South Korea. For direct contact see www.ianetwerk.nl.

                                You can call us in for networking or tailor-made service related to international science,
                                technology and innovation cooperation. The Innovation Attachés have an important
                                trendwatching role focused on science, technology and innovation developments.

                                We look forward to exploring how we can support your international ambitions. About
                                Artificial Intelligence, but also about the other top sectors and key enabling technologies.

                                On behalf of the Holland Innovation Network,

                                Juri Roerink
                                Coordinator Holland Innovation Network

                                Twitter: @ianetwerk | Linkedin: www.linkedin.com/in/ianetwerk | E-mail: ianetwerk@rvo.nl

    Colophon
    This is a publication of:                                    Quoting & referring
    Netherlands Enterprise Agency                                Please note that you are free to quote or refer to this document provided you mention the
    (Rijksdienst voor Ondernemend Nederland)                     source in your publication and inform the Central Office for the Holland Innovation Network
                                                                 at RVO.nl by sending an Email to: ianetwerk@rvo.nl

    Visiting address               Postal address                Photography and visuals
    Prinses Beatrixlaan 2          PO Box 93144                  Unsplash, Shutterstock, Dreamstime, ANP, iStock and others. We have exercised the utmost
    2595 AL Den Haag               2509 AC The Hague             care in trying to track down all rights owners of the photos etc used as visual illustration in
    T (088) 602 15 04              The Netherlands               this magazine. However, should you feel an infringement of your copyright has taken place,
    E ianetwerk@rvo.nl                                           please contact the publisher.
    www.ianetwerk.nl
                                                                 Design and print
                                                                 XeroxOBT, The Hague
    © Rijksoverheid | September 2018 | 111535
    ISSN: 1572-6045                                              Interviews, final editing, translation and overview article
                                                                 Hans van der Lee (Geschreven en Gedrukt Communicatie), Den Haag

                                                                                                                                                           AI special
Artificial Intelligence - Holland Innovation Network Special - RVO
The Netherlands                                                                                   3

The state of AI
in the Netherlands

                  The size of the Netherlands belies its economic and innovation power: the
                  131st country in the world in terms of area is the world’s 66th most populous
                  country, the 18th largest economy of the world, and ranks 3rd in the Global
                  Innovation Index. One area in which the Netherlands’ innovation power
                  bears fruit is artificial intelligence (AI). Some highlights from the sectors
                  agriculture, healthcare, infrastructure, mobility and more.

Artificial Intelligence
Artificial Intelligence - Holland Innovation Network Special - RVO
4                                                                                                                                          The Netherlands

                                                                                                                    Computer chip
                                                                                                                    design and
                                                                                                                    manufacturing is
                                                                                                                    integral to AI
                                                                                                           Bachelor, Master and Ph.D. students in
                                                                                                           informatics, artificial intelligence and other
                                                                                                           fields of study, learning to conduct research
                                                                                                           as well as entrepreneurial skills. The city of
                                                                                                           Amsterdam has pledged € 4 million; the rest
                                                                                                           will be provided by participating (multina-
                                                                                                           tional) companies and acquired from other
                                                                                                           sources.
    For over 250 years the fundamental drivers of       In machine learning, the computer ‘learns’ to
    economic growth have been technological             distinguish connections and underlying             Maarten de Rijke, Professor of Informatics
    innovations. The most important of these are        relations in data, without having been             and coordinator for the new AI hub, told
    what economists call general-purpose                explicitly programmed in advance. Machine          newspaper Het Parool: “With regard to AI,
    technologies, a category that includes the steam    learning operates through learning from data,      there is only one viable option: be active. If
    engine, electricity, and the internal combustion    building a model of reality and constantly         today you are at the top and do not follow
    engine. Each one catalysed waves of comple-         refining it. On the basis of such a model, these   through, tomorrow you will suddenly have
    mentary innovations and opportunities.              algorithms are capable of producing data-          fallen behind.” AI studies have drawn 400-500
                                                        driven predictions as outcomes.                    students over the past few years. And current
    The internal combustion engine, for                                                                    research is already world-class, says De Rijke.
    example, gave rise to cars, trucks, airplanes,      Subsequently, they continuously build upon
    chainsaws and lawnmowers, along with                these outcomes, ever refining them. A very         A good example of the AI created at the Science
    big-box retailers, shopping centres, cross-         important concept within the field of machine      Park, is startup company Scyfer. Founded five
    docking warehouses, new supply chains, and          learning is the concept of deep learning. This     years ago, it was picked up by US computer
    suburbs. Companies as diverse as Walmart,           specific machine-learning technology uses          chip giant Qualcomm in the summer of 2017.
    UPS and Uber found ways to leverage the             so-called neural networks and a hierarchy of       Scyfer’s AI will be integrated in Qualcomm’s
    technology to create profitable new business        concepts. Complex concepts are defined in          chipsets and wireless solutions. In order to
    models.                                             terms of simpler concepts. The software            foster similar startups the hotspot will create
                                                        optimizes itself without humans necessarily        preconditions conducive to joint education,
    The most important general-purpose                  being able to see how the algorithm adjusted       research and development in cooperation with
    technology of the 21st century is AI. In            itself. In other words, we only see input and      the private sector.
    particular machine learning, i.e. a machine’s       output (i.e. the quality of a prediction); the
    ability to keep improving its performance           algorithm used is a ‘black box’.                   Developing an AI institute goes beyond
    using examples (or data) without humans                                                                Amsterdam and its university, De Rijke states.
    having to explain exactly how to accomplish         The challenge with AI development, big data        “What we see, is that the societal debate on AI
    all the tasks it is given.                          and machine learning is not only technologi-       has been opened. Everyone supplies data to
                                                        cal, but also lies in new organization and         companies such as Google and Facebook. We
      The advent of AI                                  business models as well as using large             have to set our own expertise next to it. The
    Over the past few decades the calculating           volumes of data in a meaningful way.               best way to have a degree of control is to
    power of computers has increased exponen-                                                              develop technology that works and meets our
    tially, according to Moore’s Law. Big data has        Amsterdam: new world-class AI hub                own values.”
    arrived, as a result of a large part of our life    In 2019, the University of Amsterdam (UvA)
    playing out in the digital world. A third factor,   will start construction of a ‘Premier League’      Professor Max Welling, research chair in
    enabled by the first two mentioned, is the rise     AI hotspot at its Science Park. The planned        machine learning at UvA and co-founder of
    of machine learning.                                11,000 sqm building will accommodate               Scyfer, is also involved with the new hub. He

                                                                                                                                                 AI special
Artificial Intelligence - Holland Innovation Network Special - RVO
The Netherlands                                                                                                                                       5

stresses the fact that the planned new AI hub   retaining AI talent. “I hope and fully expect      government policy. Furthermore, automated
is not meant to be an exclusively UvA or        our government to recognize that the return        integration of data sources and streams will
Amsterdam operation. “We are aiming for an      on investment in AI is many times greater          lead to substantial improvements in
open consortium of academic partners with a     than the investment itself.”                       efficiency.”
national scope and are in discussion with       Until recently he would have mentioned the
various industrial stakeholders.” Welling       United States as the most important country        International, private and university partners
explains that participating companies can be    with business opportunities for the Dutch AI       in CBDS include, the Netherlands’ Central
either multinationals, SMEs or startups. The    sector. He now says the Netherlands should         Bank (DNB), TNO Innovation for Life, the
idea is for a company to establish a research   do better by also focusing on China, due to        WorldBank, Microsoft, IBM, CapGemini, CGI,
lab in the new building, thus investing in a    the size and growth rate of its AI sector.         Dell-EMC, Maryland University and Michigan
laboratory infrastructure and co-financing                                                         University.
research.                                         Statistics Office in big data vanguard
                                                In September 2016, the Dutch Central Bureau        Tjin-A-Tsoi: “We have quite a few challenges
  Dutch AI fields of excellence                 of Statistics (CBS) launched a unique initiative   ahead of us with using AI for statistics. For
Professor Welling considers the following       for big data research: the Center for Big Data     example, government entities are obliged by
subfields as sectors where the Netherlands      Statistics (CBDS). Within CBDS national and        law to provide CBS access to their data. With
has a role to play with regard to AI:           international partners from government,            regard to big data of private companies it is as
• High-tech: computer chip design and           business and science cooperate on big data         yet unclear if there is a legal obligation.
  manufacturing, ASML, NXP;                     technology and methods for producing
• Agriculture: innovative AI and robotics in    official statistics.                               Another issue is the nature of big data. Much
  precision agriculture, greenhouse                                                                of it is high-frequency machine-generated
  technology, food factories;                   Tjark Tjin-A-Tsoi, the director-general of CBS:    data, for example from sensors. Integrating
• Healthcare: excellent integrated and          “Big data has the potential of enriching,          this data stream with our existing survey and
  networked hospital structure, prevention      improving and speeding up the production of        administrative data requires new methodo-
  using social media;                           statistics, with less costs and administration.    logical solutions. ‘Translating’ big data into
• Logistics, planning: reinforcement            Many organizations are investing in big data       the statistics we would like to see is difficult
  learning.                                     technology. However, producing high-quality        in two respects. Firstly, what exactly do we
                                                statistics on the basis of big data is complex.    measure when using traffic data or social
Welling sees Canada as a model country for      This concerns both quality and continuity.         media use? Your sample is not random
the way it has set up and continues to invest   Our ultimate goal is producing better and          anymore. The population you accidentally
in AI. He specifically means educating and      more relevant statistics in support of             measure could be just a part of the popula-
                                                                                                   tion. Twitter users are not a representative
                                                                                                   cross-section of society.

                                                                                                   When using big data, often you will have to
                                                                                                   combine different data sources in order to
                                                                                                   have an accurate picture. Big data is not
                                                                                                   always of the highest quality: sensors can
                                                                                                   break down, provide false readings – factors
                                                                                                   for which you have to adjust the results.”

                                                                                                              ASML is one of
                                                                                                              the leading tech
                                                                                                              companies actively
                                                                                                              involved in Dutch AI

Artificial Intelligence
Artificial Intelligence - Holland Innovation Network Special - RVO
6                                                                                                                                     The Netherlands

                                                        AI in infrastructure: the case of               Another growth sector within AI is analyzing
                                                        VolkerRail                                      and interpreting written or spoken texts.
                                                      CQM (Consultants in Quantitative Methods),        Google Translate has been substantially
                 VolkerRail                           an Eindhoven-based company, applies data
                                                      science and AI to business opportunities in
                                                                                                        improved by using AI. This smart program has
                                                                                                        developed its own intermediate language by
           employs CQM’s                              high-tech, logistics, energy and agri & food.     which any language can be translated into
                                                      For railway infrastructure company VolkerRail     any other. At CQM, smart text analysis is used
         AI solution for the                          they implemented AI to detect irregularities      to support marketing departments in
        timely detection of                           on the railway tracks. Huub van den Broek,
                                                      lead consultant on this project: “A specially-
                                                                                                        analyzing large volumes of reviews and
                                                                                                        ratings. What are their clients talking about,
                  rail faults                         equipped VolkerRail train produces video          what is the general sentiment, what settings
                                                      images of the Dutch railway system. The           do they have to adjust to improve their
                                                      images are inspected for faults by trained        products and services?
    Where are we now, after 18 months of CBDS?        inspectors. Small defects can grow into larger
    “The organization has been structured,            problem areas. Detection is imperative.           Van den Broek sees the greatest opportunities
    people have been recruited, the finance has       Preventive maintenance ensures trains are         for ‘Dutch AI’ “in fields where the Nether-
    been taken care of, there is an impressive list   using the tracks safely.”                         lands is traditionally strong, such as agri &
    of partnerships, data have been collected,                                                          food, banking, energy, high-tech and
    priorities have been defined, and over the        Together with VolkerRail, CQM developed and       logistics. AI systems for decision-making
    next few months new project results will be       trained a so-called deep convolutional neural     support can take these sectors to the next
    published frequently.”                            network to automatically detect such faults.      level.” He identifies major tech companies as
                                                      Its network architecture is based on how the      more important than countries for determin-
    One example project concerns the energy           human brain processes visual information.         ing where Dutch AI businesses should focus.
    transition, from predominantly fossil fuels to    After ‘seeing’ thousands of examples of intact    In addition to well-known US players, these
    sustainable energy. CBS is charting energy        and faulty rails, the network has ‘learned’       are Alibaba, Baidu and Tencent. Van den
    consumption to present opportunities for          which characteristics determine classification    Broek sees the big (user) data these compa-
    decreasing it or making it sustainable.           of specific faults. When the neural network       nies have collected as their most valuable
    Optimal placement of charging stations for        recognizes such a characteristic in new video     asset for success.
    electric vehicles (EV) can be determined by       images, the inspector is warned so the
    measuring EV density, ensuring locations          necessary actions can be implemented.             AI education, training and life-long learning
    with maximum profit. Another aspect is the                                                          is the most formidable challenge for the
    particulate matter situation, in relation to      Van den Broek: “Of course, the greatest part      coming years, according to Van den Broek. In
    traffic streams intensity: how can we improve     of the railway tracks is not faulty at all. The   conclusion, he believes the positive outcomes
    both?                                             neural network allows the inspector to focus      far outweigh the negative aspects of AI. “Let
    The internet economy is another ongoing           on relevant images where suspicious spots         us not get stuck in negativism, stimulate
    research project. What is the size of it, how     are visible, increasing efficiency. A fine        entrepreneurship in a major way and let AI
    much do the Dutch buy from foreign internet       example of AI application.”                       bring our welfare and well-being to unchart-
    companies? This is a cooperation between                                                            ed levels.”
    CBS and Google.                                   Image recognition and visual inspection is a
                                                      field where AI has developed at a giant pace.       Ethics, norms and values:
    Cities in the Netherlands are researching the     Self-learning systems now outperform                the Dutch approach
    smart city concept. CBS is assisting them with    humans. The private sector employs a large        Former lawyer Bendert Zevenbergen is a
    its urban data initiatives, providing methodo-    number of people for visual inspection,           research fellow on the ethics and politics of
    logical support with uniformly structured         whereas this task can efficiently be ‘out-        AI at Princeton University. Working at the
    data.                                             sourced’ with AI.                                 Center for Information Technology Policy and

                                                                                                                                             AI special
Artificial Intelligence - Holland Innovation Network Special - RVO
The Netherlands                                                                                                                                         7

the Center for Human Values, he views AI           is smart mobility: vehicles connected to the
through the prism of norms and values.             mobility infrastructure and to each other
                                                   through internet. The connected vehicle is so
“I believe the Netherlands has a special role to   much more than just an autonomous /
play in terms of the ethical dimension of AI.      self-driving car. Steinbuch means vehicles and
We have always been a country that is              infrastructure communicating in real time,
relatively strong in societal and ethical          for example with cars platooning on the road.    mobility? Well, our soccer robots use the
reflection on technological issues. What we                                                         same algorithms which can be used for smart
should do is acquire the knowledge from            “The classic concept of mobility has come        cars. The robots have spatial awareness and
various AI sectors and test it in AI applica-      under pressure, because of climate change        learn from what is happening on the ground,
tions, subsequently feeding the world our          and the transition from fossil fuels to          as an autonomous car will do when driving
ethical and societal conclusions.”                 sustainable energy. Traffic congestion in        on the road. Is that a lamp post or a slender
                                                   metropolitan areas is a global problem. If we    student crossing the road for which it has to
In Zevenbergen’s analysis, both China and the      can find ways to apply technology for making     stop?”
US are not the kind of society the Netherlands     cars smarter and more autonomous, the
can be or wants to be. Therefore, building a       drivers can perform tasks other than actually    The Netherlands is the first country in the
cooperation network with the countries             driving. In other words, we can make better      world where public roads have been used to
around us seems logical: France, Germany           use of our valuable time.”                       test connected mobility. Steinbuch’s
and the United Kingdom. “The French                                                                 university has cooperated with TNO and DAF
government has recognized that vast                Steinbuch sees many more applications for        Trucks to practice twin driving with trucks.
amounts of money have to be invested in            smart mobility such as car-sharing, reclining
order to advance French AI. The same goes for      bicycles, NS hire bikes – all as part of one     “Most automotive companies’ R&D focuses
Germany, where the automotive sector has           smart mobility concept. Computer chips           on self-driving cars, not on connected
also produced a standard textbook in the first     producer NXP already provides most of            driving. Our leading position is an excellent
manual for autonomous cars, which is now           automotive electronics for all major car         opportunity to build upon”, he concludes.
used internationally.                              brands, thereby having a unique position in
                                                   the world.                                         AI in agriculture
  Developments in connected mobility               Robotics also comes into play. His group has     Erik Pekkeriet is senior business development
Maarten Steinbuch is Professor in Control          won the world soccer championship for            manager Agro Food Robotics at Wageningen
Systems Technology at the Eindhoven                robots three times. “You may ask yourself        University & Research. What can he tell us
University of Technology. One of his projects      what does this have to do with smart             about the AI strengths of the Netherlands in
                                                                                                    agri & food? “There are too many fields to
                                                                                                    mention where AI has been successfully
                                                                                                    applied. For example in feature extraction,
                                                                                                    classifying food quality, pest and disease
                                                                                                    recognition as well as harvest and crop
                                                                                                    handling using vision technology (either in
                                                                                                    2D/Slam or 3D (TOF, Stereo, structured light,
                                                                                                    VI). Actual challenges now are time-restricted
                                                                                                    learning, to monitor and control crop progress
                                                                                                    (season, cultivar, lighting, grower-specific crop
                                                                                                    handling and greenhouse or field distribution)
                                                                                                    with the broad variety on plant sensors, as well
                                                                                                    as human robot collaboration. Wageningen
                                                                                                    University & Research is involved in all of these
                                                                                                    technologies.” Priority countries for the Dutch
                                                                                                    AI sector in his field are Australia, Japan and
                                                                                                    the US.

Artificial Intelligence
Artificial Intelligence - Holland Innovation Network Special - RVO
8                                                                                                                       United States

                                                   Preparing
                                                   the Future
                                                   of Society
       The recent surge in investments and strategic focus on Artificial Intelligence
       (AI) is based on serendipity: many favorable conditions coming together at
       the same time. This article on developments in the United States in general
       is complemented by separate articles on AI developments in Boston and
       Silicon Valley.

    By Sigrid Johanisse, Holland Innovation Network, Washington, D.C.   with China following in second place. [McKinsey’s State of
                                                                        Machine Learning and AI, 2017].
    Chips have become faster, cheaper and highly capable of
    processing vast amounts of data; the exponential growth of          According to CB Insights, the majority of all venture
    data via larger platforms; the rise of tech talent which is able    capital-funded AI startup companies are still in the US,
    to develop algorithms; low costs of data storage; availability      with most of them located in San Francisco and Silicon
    of capital; and the advent of new areas for commercial              Valley. New York and Boston come in third and fourth.
    application. From autonomous driving to agriculture and             The average starting salary of a computer science graduate
    cyber security, from defense to fintech or healthcare: AI can       who can develop algorithms is between USD 150,000-
    be the game changer and everybody wants to be first.                200,000.

    In view of these conditions and areas for application, it is          Support by US government
    not surprising to see that AI and machine learning have             It is not just the private sector which is embracing AI.
    become the next investment boom of the tech industry in             The US government is also supportive. In order to ensure
    the United States (US).                                             global leadership in AI by the US, the White House has
                                                                        chartered a Select Committee on AI under the National
    Over the last five years, companies such as Apple, IBM,             Science and Technology Council. It includes the National
    Microsoft, Google, Facebook, Intel, Amazon, Salesforce,             Science Foundation (NSF), the Office for Science and
    Uber et cetera have invested billions in AI research (90%)          Technology and DARPA. The NSF and DARPA have put
    and acquisitions (10%). US-based companies have                     forward machine learning and robotics as one of the main
    absorbed 66% of total worldwide AI investments in 2016,             strategic research areas. In September 2017, the White

                                                                                                                            AI special
United States                                                                                                                                            9

                                                                    AI hub New York
                                                                    New York is the third AI hub in the US. With challenging AI startups such as X.AI,
                                                                    Clairifai, AlphaVertex, Behold.AI, CambrianIntelligence and Klustera the city sets
                                                                    the pace and influences business models in fintech, healthcare, robotics and
                                                                    consumer services. These companies specialize in image and voice recognition,
                                                                    but also perform highly specialized tasks in manufacturing. Facebook’s AI
                                                                    department is anchored in NYC, as well as IBM’s global Watson headquarters.
                                                                    Columbia University has a data science institute and the Technion-Cornell
                                                                    Institute is the bridge between New York and Tel Aviv. In October 2017, Dutch
House published its vision on the future of technology in           startup incubator B. NYC was established to provide AI-focused startups from
the US, stressing the importance of Machine Learning and            the Netherlands with a (speedy and) soft landing in the US.
autonomous technologies for transport and other areas
which would put America First.
                                                                  en focus of AI, which ignores the impact and the diverse
To prepare the workforce for the jobs of the future, the          reality in which we live.
Trump administration will invest strongly in STEM
education and workforce training programmes. In                   Under Obama’s administration [Artificial Intelligence, Automa-
addition, the US Congress has adopted federal legislation         tion and the Economy, 2016] a multistep approach was being
to facilitate autonomous test driving throughout the US.          developed to deal with these societal policy-related issues.
This has attracted investments from Volkswagen, BMW               These efforts are now being continued at the Center for
and Mercedes to the US in order to keep up with the latest        Information Technology Policy (CITP) in Princeton, under
technological developments.                                       the guidance of Ed Felten. Via the Princeton Dialogues of
                                                                  AI and Ethics, the CITP publishes case studies combining
  Societal debate in the US                                       various scientific disciplines such as ethics, law, computer
At the same time, is not only commerce, science and               science and policy-making.
government who are paving the way towards the future.
AI and machine learning have sparked a broad societal             Industry has also joined forces by funding the not-for-
debate regarding their impact on our future. Will robots          profit OpenAI Institute. In it, large tech platforms openly
eventually ‘compete’ with humans or take our place and            share knowledge and debate the impact of AI. Stanford
jobs? Do we need to develop a conditional framework in            University is writing history by studying the effects of AI
which AI is only there to assist people and improve the           on our life and work for the next 100 years (AI.100).
quality of life? What are the legal and etthical aspects of AI?   Stanford professor FeiFei Li and Melinda Gates have
                                                                  started an organization AI4ALL to eliminate bias from
There is widespread concern in the US Congress that AI ‘is        algorithms. The AINow Institute in New York brings
taking away jobs’. Several organizations are questioning          together several disciplines to understand the social
the ethics - or better the lack of ethics - and the tech-driv-    implications of AI.

Artificial Intelligence
10                                                                                                                                        United States

                                                                              AI fight club
                                                                              In the US, the best and brightest minds are routinely
                                                                              challenged through competitions to develop novel
                                                                              technologies which are potential game changers in
                                                                              different fields of research and applications. Especially
                                                                              regarding the potential impact of AI and the future of
                                                                              cyber security and cyber warfare. Many such
                                                                              challenges are being organized by both government
                                                                              and industry like DARPA and Google to explore new
                                                                              areas of research in AI related to cyber security.
                                                                              Recently, a five-month contest started, run by Kaggle
                                                                              (Google), where offensive and defensive AI algo-
                                                                              rithms are doing battle to gain insight how to harden
                                                                              machine-learning systems against future attacks and
                                                                              study their limits. The results will be presented at a
                                                                              major AI conference later this year in the US.

               The AAAI (Association for Advancement of AI, founded in      Montreal and the Vector Institute in Toronto at its core.
               1979) studies all developments and shares all available      Canada is convinced the fastest way forward is having top
               knowledge. In this way, the US is not only the commercial    talent in your institute, which automatically attracts more
               testbed of innovative AI-driven solutions but also the       talent, because ‘smart people want to work with other
               societal testbed. See the debate between Elon Musk (Tesla)   smart people’ and consequently draw large investments
               and Mark Zuckerberg (Facebook) on the question whether       from US tech companies such as Facebook and Uber. The
               AI would be a fundamental risk to human civilization, and    strategy requires companies not only to invest, but also to
               where Musk called for pro-active regulation.                 open up an R&D lab in Canada where staff can both work
                                                                            for universities as well as industry. It is the elements of
                 Developments in Canada                                     knowledge exchange and interdisciplinary thinking which
               Just across the border in Canada, the Trudeau government     make all the difference. This concept opens up a lot of
               has published a Pan-Canadian Strategy on AI based on         creativity and fosters cross-overs. The Canadian govern-
               attracting talent and investments across the globe, with     ment has invested CAN $ 125 million (USD 97 million).
               the Canadian Institute for Advanced Research (CIFAR) in      The private sector contribution equals this amount.

     Al in US agriculture
     By Karin Louzada - The application of AI in agriculture promises       environmental data to determine which crop breeds and condi-
     automatic predictive solutions for making real-time decisions in       tions will produce the highest yields. Exciting applications are crop
     the field. Producing high-quality food while at the same time          imaging, pest imaging and identification, sensor-based intelligent
     reducing the use of land, water and labor resources, with 24-hour      irrigation, vertical farming and autonomous tractors and drones.
     uninterrupted production. However, its implementation comes
     with significant challenges. Agriculture is notoriously complex: an    Automated crop management solutions for thinning, weeding and
     open system with continuously changing variables, unpredictable        picking are of particular interest in California for replacing
     environmental conditions, diseases and pests, and few                  back-breaking work in a sector already suffering from a rapidly
     measurement points (e.g. yield measurements once a year). It is        dwindling labor supply. A prime example is the LettuceBot by Blue
     difficult to statistically quantify the results of intervention and    River Technology, a Silicon Valley-based company developing a
     solutions are hard to test, validate and implement.                    suite of AI technologies for agriculture. Recent conversations with
                                                                            growers and specialists in the field have taught us that most
     Nevertheless, clear opportunities for AI and machine learning have     growers are slow to take up new technologies on their own.
     been identified in, for example, accelerating the breeding process.    However, large agribusiness firms are showing great interest, as
     Seed companies are already using AI to combine phenotype data          testified by the recent purchase of Blue River by John Deere for
     (acquired through multispectral imaging) with genetic and              USD 305 million!

                                                                                                                                             AI special
United States             11

Artificial Intelligence
12                                                  United States - Boston

     Boston
     Because of its many top-ranked universities and
     research institutes, active startup scene, venture
     capital availability and the presence of mayor
     players such as IBM, Google, Boston Dynamics and
     GE, Boston is well-equipped to be a strategic hub
     for the new wave of Artificial Intelligence (AI).

                                                                AI special
United States - Boston                                                                                                                                13

                                                     The eight internationally ranked universities of the Greater
                                                     Boston area
                                                      University                              Computer Science program   AI program / research
                                                      Massachusetts Institute of Technology              X                           X
                                                      Harvard University                                 X                           X
                                                      Boston University                                  X                           X
By Walter de Wit, Christian Cohen, Tyrone Pater,      Tufts University                                   X                           X
Holland Innovation Network, Boston                    Boston College                                     X
                                                      Northeastern University                            X                           X
Boston is home to both Harvard University
                                                      Brandeis University                                X
and the Massachusetts Institute of Techno­
logy (MIT). In addition, the Boston region has        University of Massachusetts – Boston               X                           X
six other internationally ranked universities.

The Harvard John A. Paulson School of              reputation for making a global technological      in collaboration with the Berkman Klein
Engineering and Applied Sciences (SEAS) is         impact in the world. The Computer Science         Center for Internet and Society at Harvard
the engineering school which offers degrees        department offers master’s and doctoral           University.
in those fields. AI is part of its Computer        programs with courses in areas such as
Science curriculum. Courses on the theme are       algorithms, information theory, multicore         In October 2016, MIT announced the creation
widely covered, ranging from Machine/Deep/         programming and robotics. The Boston Globe        of The Engine, a new venture for entrepre-
Reinforcement Learning, Visual Recognition,        reported that MIT’s Introduction to Machine       neurs working on technological innovation
Language and Logic, and Super Intelligent          Learning has become one of the most               with the potential of changing the future. The
Machines.                                          popular classes on campus, with 700 students      Engine prioritizes breakthrough ideas over
                                                   signing up this year.                             early returns by providing capital to tough
The Harvard Innovation Lab (i-Lab) is a                                                              tech startups. On September 19th, 2017 The
resource for Harvard students interested in        The largest research laboratory at MIT is the     Engine announced it had raised USD 200
entrepreneurship. Several AI startups have         Computer Science and Artificial Intelligence      million to invest in startups. The AI startup
originated from the i-Lab: Intelligence            Laboratory (CSAIL). CSAIL’s research groups are   iSee is one of the first companies in their
Associate, Heartstring, Evisort and Finfox.        organized into three focus areas: artificial      portfolio.
                                                   intelligence, systems and theory. In addition,
The Electrical Engineering and Computer            the MIT Media Lab is participating in a USD 27      Healthcare
Science Department (EECS) at MIT has a solid       million initiative on AI ethics and governance,   The combination of data and AI has the
                                                                                                     potential to improve outcomes and reduce
                                                                                                     costs within healthcare, by applying machine
  MIT-IBM Watson AI Lab                                                                              learning algorithms and predictive analytics
                                                                                                     to reduce drug discovery times, provide
  In September 2017, IBM and MIT announced that IBM is investing USD 240 million to                  virtual assistance to patients, and diagnose
  create an AI lab in partnership with MIT. The new lab will be mobilizing over 100 AI               ailments by processing medical images,
  scientists, professors and students to carry out joint research at the IBM’s research lab in       among other things. In Boston both IBM and
  Cambridge. Researchers at the lab will pursue research in fundamental AI which include             General Electric (GE) aim to become a leader
  the research areas of AI algorithms, AI hardware and software. It also aims to increase            in AI and healthcare.
  the impact of AI on industries such as healthcare and cyber security.
                                                                                                     As mentioned, IBM recently announced that it
  The MIT-IBM AI lab is a new addition to a multi-year collaboration between MIT and                 partnered with MIT to create a USD 240 million
  IBM. In 2016 IBM announced a long-term collaboration with MIT’s Department of Brain                AI lab, with healthcare as one of the focus
  and Cognitive Sciences to make progress in the field of machine vision. Moreover, IBM              areas. Moreover, IBM spent over USD 4 billion
  invested USD 50 million in a five-year research collaboration on AI and genomics with              in the past two years in acquiring healthcare-
  the Broad Institute of MIT and Harvard.                                                            computing companies and expanding its AI
                                                                                                     capabilities. IBM acquired AI-healthcare

Artificial Intelligence
14                                                                                                                                   United States - Boston

       GE and Partners Healthcare
       GE and Partners Healthcare launched an initiative to employ                health, and other areas of medicine. Executives from GE, one of
       artificial intelligence to improve medical care. The ten years long        the nation’s largest corporations, and Partners, Massachusetts’
       effort will include clinical and technology experts at the Partners-       largest not-for-profit hospital network, said such technology has
       owned Massachusetts General and Brigham and Women’s                        the potential to help care providers do their jobs more efficiently
       hospitals working alongside engineers and developers at GE. The            so that patients receive more accurate diagnoses and better
       partnership is initially focused on developing deep learning               treatments. Without disclosing specifics, both companies said
       algorithms that analyze medical images to help identify patients           they will spend a significant amount on the initiative. And both
       that need follow-up or intervention by doctors. But over time,             stand to gain revenue if they are successful in creating useful
       they also want to create applications for genomics, population             software programs that can be sold to hospitals around the globe.

     companies Ann Harbor, Merge Healthcare,             create better standards for procedures and        working on an intelligent sepsis monitoring
     Phytel and Explores in the last two years. IBM      treatments. Welltok is an AI startup located at   system to continuously identify and risk-
     claims that Watson software is used in a dozen      Burlington, MA, and is one of the five most       stratify patients at increased risk of sepsis.
     of the largest life sciences companies and that     well-funded AI companies in healthcare (CB        Massachusetts General Hospital is partnering
     more than 55 hospitals and healthcare               Insight). Its mission is empowering consum-       with Nvidia to use its new deep learning
     organizations implemented its oncology tools.       ers to achieve and sustain their optimal          supercomputer and improve the detection,
     Watson says that it can quickly sift through sets   health. Their CaféWell Health Optimization        diagnosis, treatment and management of
     of data to perform drug research and to make        Platform connects consumers with available        diseases.
     diagnoses. Watson software analyzed 1,000           and relevant resources, benefits and rewards
     cancer patient cases and suggested potential        using a data-driven, personalized approach.         Robotics
     treatments that doctors were not able to            Some other AI startups in Healthcare in the       Another very lively AI domain in Boston is
     identify in about 30% of the cases.                 Boston region are Jibo, Legionarius,              robotics. The various universities in the
                                                         Neurosleeve, and Bergh Health.                    Boston region ensure that a significant
     Many of the Boston region-based AI startups                                                           amount of talent is available to deal with the
     have a focus on healthcare. Already men-            Boston hospitals are also looking to improve      hard problems that are often associated with
     tioned is the Cambridge-based startup               hospital care by integrating AI in their          robotics. Additionally, the sheer number of
     PathAI. Another startup located in Cambridge        organizations. The Beth Israel Medical            universities in the region results in a large
     is OM1; they use machine learning algorithms        Deaconess Medical Center (BIDMC), one of          number of startups focused specifically on
     and the vast amount of healthcare data to           Harvard’s teaching hospitals, is currently        robotics.

                                                           The Boston startup scene
                                                           Boston has a rich ecosystem of AI startups, focusing on a wide range of applications of
                                                           AI, some examples are Interactions, Indigo and PathAI.

                  Boston is                                Interactions provides virtual assistant software that combines speech recognition and
                                                           human intelligence. It is the best-funded AI startup in Boston with USD 167 million
       well-positioned to                                  funding raised. Indigo, which raised USD 150 million, helps farmers to make sturdier

              lead the next                                crops by applying machine learning algorithms to a large database of plant genomes.
                                                           The Cambridge-based PathAI developed a platform that provides automated and
               revolution in                               data-driven pathology analysis.,

     artificial intelligence                               In 2016 venture capitalists in the US invested in 658 AI-related startups, with a total
                                                           invested amount of USD 5 billion. An increase of 61% compared to 2015. The growing
                                                           interest of venture capital firms in AI is reflected in Boston. Boston-based venture
                                                           capital firm Glasswing emerged to exclusively invest in AI-related firms. In June 2016, it
                                                           announced that it aimed to raise USD 150 million to invest in AI startups. Other AI
                                                           investors based in the Boston area are Hyperplane investors, Procyon Ventures and GE
                                                           Ventures.

                                                                                                                                                 AI special
United States - Boston                                                                                                                               15

The MIT Duckietown course is a class on
advanced autonomy at MIT. It is a hands-on,         Boston AI events
project-focused course and involves self-
driving vehicles and high-level autonomy.           AI World Conference & Expo, December 3- 5, 2018
Students are asked to design an autonomous          The industry’s largest independent event focused on enterprise AI and machine
robo-taxi system for the city of Duckietown.        learning. Designed to help business and technology executives cut through the hype,and
Although they work on scale, the problems           learn how advanced intelligent technologies are being successfully deployed.
students encounter are comparable to the
problems of Tesla and Uber. All materials,          Boston AI Meetup – monthly events
such as hardware design, software, and              The Boston AI Meetup consists of almost 2000 members and have monthly meetups
teaching materials, are released as open            about the most exciting and inspirational AI innovations in the Boston region.
source.

Furthermore, the prestigious Worcester            create and scale the next generation of
Polytechnic Institute (WPI) is making strides     robotics companies. They prove entrepre-
when it comes to robotics. With the               neurs and startups workspace and resources                Overall, Boston
first-of-its-kind bachelor’s degree in robotics
engineering and one of the first such
                                                  they need to develop, prototype, test, and
                                                  commercialize their products and solutions.
                                                                                                            seems ready to
graduate programs in the nation, WPI is at        Another program is Autodesk BUILD, their                  become one of the
the forefront of robotics education and           program supports research and startup teams
research.                                         as they build and test projects and proto-                leading AI hubs
                                                  types. Residents can access training,
In total, there are almost 40 research labs in    expertise, areas for large-project work and
the Boston region that have a focus on            more than 60 pieces of advanced fabrication      known for their collaborative robots Sawyer
robotics. Some other notable examples are         machinery.                                       and Baxter. In addition, there are dozens of
MIT CSAIL, Harvard Wyss Institute, Harvard                                                         robotics startups in the Boston region,
Biodesign lab, the Robotics and Intelligent       Boston is home to several leading robotics       ranging from warehouse and agricultural
Vehicles Research Laboratory at Northeastern      companies. Originally a spin-off of MIT,         automation to exoskeletons and social
University and the New England Robotics           Boston Dynamics builds advanced robots           robots.
Validation and Experimentation Center at the      highly skilled at mobility, agility, dexterity
University of Massachusetts.                      and speed. Their robots use sensor-based         Overall, Boston seems to be ready to become
                                                  controls and computation, which allows           one of the leading AI hubs. Boston region’s
In addition to the various research labs, there   capabilities of complex mechanisms. iRobot       unique position in enabling new technolo-
are other initiatives that help to improve the    is a leading global consumer robot company,      gies with its leadership in academic and
robotics ecosystem in Boston. Located at the      which designs and builds robots that             medical research resulted in a robust and
Seaport area, MassRobotics is the collective      empower people to do more both inside and        growing AI ecosystem. Combining this with a
work of a group of Boston-area engineers,         outside of the home. They seek to build an       very active startup scene and industry
rocket scientists and entrepreneurs, who          ecosystem of robots and data to enable the       heavyweights such as Amazon, IBM Watson
created an innovation hub and startup             smart home, both through internal initia-        and General Electric (re)locating to the state,
cluster focused on the needs of the robotics      tives and by building strategic external         Boston is well-positioned to lead the next
community. MassRobotics’ mission is to help       partnerships. Finally, Rethink Robotics is       revolution in Artificial Intelligence.

Artificial Intelligence
16                                                                                                                         United States - Silicon Valley

     Silicon Valley

     In 2017, the US was the second largest investor in AI companies with 38% of total worldwide
     AI investment. The US has now been surpassed by China with 48%. In the number of AI
     companies the US is still in the lead. Based on keyword searches by CB insights, China is also
     taking over the majority of AI patent applications.

     By Wouter Roelofs, Holland Innovation Network,   In Silicon Valley AI is a top priority, with     Most investments, advancements and
     San Francisco                                    Google in the lead. Since 2012 it has acquired   applications are now made in the Healthcare,
                                                      the largest number of AI startups, followed by   Cybersecurity and Enterprise AI areas. The big
                                                      Apple, Facebook and Amazon. At the annual        shift in last year’s developments was from AI
                                                      Google I/0 conference in March 2018, they        in the cloud capabilities to AI to the ‘edge’,
                                                      showed their most advanced conversational        meaning to devices that we use every day,
                                                      AI (chatbot) interacting with humans and         such as our smartphones (iPhone X with
                                                      understanding more context than ever             facial recognition) and other devices
                                                      before.                                          executing more complex tasks.

                                                                                                       Questions most often heard are: how can AI
                                                                                                       enhance our products to the next level? But
                                                                                                       also: is AI the blessing in the tech world we
                                                                                                       were all waiting for or are we ‘summoning the
                                                                                                       demon’? Or will it ‘only’ eliminate our jobs?

                                                                                                       AI is already benefiting platforms and
                                                                                                       businesses such as Uber, Netflix and many
                                                                                                       others. AI is transforming everyday products to
                                                                                                       be more convenient for users and will likely
                                                                                                       transform society as a whole. Companies such
                                                                                                       as IBM, Intel and Salesforce have business
                                                                                                       solutions to integrate AI technologies in
                                                                                                       existing work processes and making businesses
                                                                                                       more efficient and effective.

                                                                                                       Amazon has highly automated warehouses
                                                                                                       with robots (100.000+) but the more refined

                                                                                                                                               AI special
United States - Silicon Valley                                                                                                17

                                                 Large companies in the western United States
                                                 are even setting up AI-focused venture capital
                                                 funds to solely focus on AI, such as Google’s
                                                 Gradient Venture, Microsoft Ventures and
                                                 Salesforce Ventures. This indicates the
                                                 importance of AI to US companies.

                                                 When looking at the fifty largest startups in
                                                 AI, it is clear that Business Intelligence and
                                                 Analytics and Core AI are dominating the
                                                 invested industries in the United States.
                                                 However, in the rest of the world, Healthcare
                                                 received the most investments. Companies
                                                 are not only investing in startups but also in
and delicate tasks are still performed by        patents. CB Insights analyzed five of the
humans. AI replaces dull, dangerous and dirty    largest tech companies: Google, Microsoft,
jobs but a whole realm of new jobs will also     Amazon, Apple and Facebook. Seattle-based
emerge.                                          company Microsoft leads the group with
                                                 more than 200 AI-related patent applications,
Furthermore, a debate has started on the         followed by Google, Amazon, Facebook and
development of logarithmical biases, as most     lastly Apple. Even though American compa-
developers are young males, white and Asian.     nies only really started investing in AI in 2012,
Initiatives as AI4ALL from Stanford University   it is evident that AI will continue to grow
pledge for more diversity in tech and to         exponentially in the near future.
involve a wider diverse group in the develop-
ment and the capabilities of AI.                   Research, innovation and talent
                                                 In Silicon Valley, there is a shortage of talent;
  Trends and investments                         with thousands of unfulfilled open applica-
Since 2012, venture capital firms have           tions at all major companies. Also, for AI a
engaged in escalating levels of investment       whole new workforce is needed. Some of the
activity to AI start-ups. The trend reached an   new professionals are educated and trained at
all-time high in 2017: in the first six months   the Californian universities, but talent is         In Silicon Valley
alone venture capital participated in USD 1.8    scouted worldwide.
billion worth of disclosed equity funding        All major universities in California have           there is a shortage of
across 88 deals to private AI companies,
already exceeding 2016’s funding total.
                                                 research groups on AI and experiment with
                                                 application of research. The universities and
                                                                                                     talent, which is
                                                 corporates work together on a wide variety          scouted worldwide
Silicon Valley makes up for 40% of invest-       of projects. Nowadays, main research goes
ments in AI globally. This may not come as a     into extracting quality data with vision
surprise, with the Bay Area housing the          analytics and analysing existing data. With
headquarters of some of the world’s largest      Machine Learning the focus is on getting
tech companies such as Google, IBM and           computers to learn quicker with less data
Apple. Since 2012, Google has acquired           input.
twelve AI startups, making it the most active
acquirer. Over the past five years, invest-      Two examples of major universities in
ments in AI have increased significantly. In     California are Stanford and UC Berkeley:
2016, McKinsey estimated global invest-          Stanford University hosts the ‘One Hundred
ments in AI to be between USD 20 and 30          Year Study on Artificial Intelligence (AI100)’
billion, an increase of 300% since 2013.         and invites leading thinkers from several

Artificial Intelligence
18                                                                                                 United States - Silicon Valley

                          institutions to study the long-term implica-        focusing on 3D model recognition issues with
                          tions of artificial intelligence in all aspects     3D prediction learning modelling. Another
                          of life.                                            angle that they study is the humanities
                                                                              perspective of ‘value alignment problems’,
       The ethical and    At the Stanford ‘Artificial Intelligence Lab’ and   how humans interact with computers and the
                          ‘The Vision Lab’, researchers focus on              difficulty to state specifically what ‘we’ want
     legal implications   computer and human vision. They learn               and how to learn from that to develop safer

        of AI are being   computers to analyse images, understanding
                          patterns, get meaningful interpretations and
                                                                              designs for AI.

               debated    develop intelligent algorithms. Further               Debate on ethics and regulation
                          research is done in object recognition, scene       Scientists and entrepreneurs are debating the
                          categorization, integrative scene understand-       implications of artificial intelligence. Research
                          ing, human motion recognition, material             and tech industry heavyweights have very
                          recognition and the underlying neural               different views on AI. The late Stephen Hawking
                          mechanisms that enable the human visual             (University of Cambridge), Elon Musk (Tesla,
                          system to perform high level visual tasks with      OpenAI, Neuralinks, SpaceX, Hyperloop,
                          speed and efficiency.                               Paypal) and Bill Gates (Microsoft) have
                                                                              repeatedly warned about the dangers of IA,
                          The Vision Lab is headed by associate               describing it as a potential threat to human
                          professor Fei Fei Li. She addresses the need        existence. Others, like Mark Zuckerberg
                          for education in AI (with the AI4ALL program)       (Facebook), Larry Page (Google) and Ray
                          and promotes diversity and fair representa-         Kurzweil (Singularity University) are very
                          tion of women in the field to get to better AI      optimistic about AI and see a bright future.
                          results.                                            Some even see a future with seamless integra-
                                                                              tion of humans and robots into a new species.
                          UC Berkeley offers the ‘Berkeley Artificial
                          Intelligence Research’ (BAIR) program. In this      Recently, a Stanford University study on AI
                          program over 100 (student) researchers are          sparked the growing debate in AI ethics and

                                                                                                                       AI special
United States - Silicon Valley                                                                                                                           19

regulation. Two researchers experimented on       build things and the world gets better. But         want to create a new kind of AI lab, operating
how artificial intelligence could identify        with AI especially, I am really optimistic. (…)     outside the control of the large corporates
people’s sexual orientation based on one’s        AI is going to deliver so many improvements         and release AI knowledge to the public.
face with an 80% accuracy. The researchers        in the quality of our lives.”. Elon Musk
defended their work by saying that they had       responded the next day on Twitter: “I’ve            Essentially, OpenAI is a research lab meant to
studied existing technologies that are already    talked to Mark about this. His understanding        counteract large corporations who may gain
widely used by companies and governments,         of the subject is limited.”                         too much power by owning super-intelli-
and that they wanted to alert the public about                                                        gence systems devoted to profits, as well as
its dangers.                                      A debate is going on about how to develop and       governments which may use AI to gain power
                                                  control AI. Should it be developed open and         and even oppress their citizenry. One of the
According to ethicist Jake Metcalf of Data &      accessible to all or in the contrary should it be   advantages of a totally open program is that
Society, more social scientists are using AI      developed in a closed academic and corporate        the labs can collaborate with anyone because
intending to solve society’s ills, but they do    environment, only accessible to experts until       they can share information freely, discovering
not have clear ethical guidelines to prevent      we understand how to control AI?                    and enacting the path to safe artificial general
them from accidentally harming people. This                                                           intelligence. OpenAI conducts fundamental,
controversy shows the need to develop               Open AI                                           long-term research toward the creation of
standards for the use of AI and data science.     In October 2015, the start-up OpenAI was            safe AGI.
                                                  announced by Elon Musk and Sam Altman
A debate between tech titans Elon Musk and        (Y combinator). Together with investors they        Many tech companies such as Amazon, Micro-
Mark Zuckerberg triggered the debate on AI        set up OpenAI pledging over USD 1 billion.          soft and LinkedIn cooperate with OpenAI and
regulation. Elon Musk pushed for the              OpenAI is a not-for-profit artificial intelli-      are contributors or project sponsors.
proactive regulation of artificial intelligence   gence research company that aims to
at a U.S. governors meeting: “I keep sounding     promote and develop friendly and safe AI in
the alarm bell, but until people see robots       such a way as to benefit humanity as a whole.
going down the street killing people, they        The organization aims to ‘freely collaborate’
don’t know how to react, because it seems so      with other institutions and researchers by
ethereal”. Mark Zuckerberg, disagrees with        making its patents and research openly
these doomsday predictions: “I think you can      available to the public. With OpenAI they

Artificial Intelligence
20                                                                                                China

     Brain-inspired
     Artificial Intelligence
     in China By Taake Manning, Ma Qing and David Bekkers, Holland Innovation Network, China

              Saudi Arabia is rich in oil and able to supply a large part of
              the world’s demand. Similarly, China has the potential to
              become the world’s largest supplier of Artificial Intelligence
              (AI) products.

                                                                                               AI special
China                     21

Artificial Intelligence
22                                                                                                                                                        China

     Consider the inputs needed for AI: access to
     capital; cheap, easy-to-get (cloud) computing
     power; a large pool of university-trained IT          Table 1: largest research institutes in brain-inspired AI
     specialists; and, above all, lots and lots of         In comparison, the largest institute in Europe is Imperial College London with 165
     data. China has it all.                               publications.

     The volume of data is particularly striking.            University / Research Institute         Publications Main Department
     China has at least 750 million mobile internet
                                                             Institute of Automation CAS                    416     Research Center for Brain-inspired
     users eager to experiment and explore the
                                                                                                                    Intelligence
     possibilities that a connected life offers. Most
                                                             Huazhong UST                                   381     State Key Lab of Image Processing and
     of them use apps belonging to the three
                                                                                                                    Intelligent Control
     largest ecosystems: Tencent, Alibaba and
     Baidu. Also, the government and public                  Southeast University                           342     State Key Lab of Computer Network and
                                                                                                                    Information Integration
     services are avid collectors of information.
                                                             Harbin IT                                      292     School of Computer Science; Depart-
                                                                                                                    ment of Computer Science Shenzhen
       Large investments in AI                                                                                      Graduate School
     China invests heavily in the development of
                                                             Tsinghua University                            257     Tsinghua National Laboratory for
     AI technologies and products, and with good
                                                                                                                    Information Science and Technology
     reason given the societal challenges it faces.
                                                             South China University of Technology           245     School of Computer Science
     The country has a population ageing at an
     unprecedented rate. At the same time, its               University of Electronic S&T                   240     State Key Lab of Neuro-information
     education system is struggling to train young
     people in the volume and to the level and
     skills needed. China is a middle-income
     country with the incidence of non-transmitta-
     ble diseases of a rich one, especially lifestyle-   patient’s or elderly person’s behaviour. China           Large-scale research projects
     related diseases. It is a rapidly urbanizing        expects AI technology to play an important           To develop enabling technologies, large-
     society but it struggles to keep people and         role in the country’s industrial development         scale research projects will be set up. These
     goods moving in its cities. AI can play a role      as well. For example, it aims to be a smart          are called National Key AI Projects, with
     in meeting these challenges, for instance           industry and intelligent vehicle superpower          focuses on big data, cloud computing, brain-
     with better diagnoses and monitoring of a           by the year 2025.                                    inspired AI, quantum-accelerated machine
                                                                                                              learning as well as brain sciences. Policy
                                                           Chinese AI policy                                  support also includes earmarked funds,
                                                         The Chinese government is creating a policy          talent training and an improved business
                                                         framework to foster a healthy environment            environment. In parallel, the government is
                                                         for AI development. Main building blocks are         taking measures to reduce the impact of AI
                                                         a 15-year long-term national policy paper and        on the labour market as well as addressing
             In terms of quality                         a 3-year short-term Action Plan.                     the legal and ethical challenges this new
                                                                                                              disruptive technology implies.
            of AI science, China                         In the long run, China aims at catching up

            is on a par with the                         with world leaders in AI technology and
                                                         applications by 2020. It expects to make
                                                                                                              In the near term, the 3-year Action Plan sets
                                                                                                              targets in more detail for a number of AI
                  world average                          breakthroughs in basic research on AI                products and applications, such as intelligent
                                                         theories by 2025. The final goal is for China to     connected vehicles, robots and AI-supported
                                                         become a ‘premier global AI innovation               medical diagnosis. Smart sensors, neural
                                                         center’ in 2030. In quantitative terms, the          network chips and open source platforms
                                                         15-year policy paper envisages China’s core AI       will be supported to form AI infrastructure.
                                                         industries reaching 1 trillion Yuan (USD 159         Measures will also be implemented to build
                                                         billion), driving surrounding industries to          industry clusters, key laboratories and
                                                         exceed more than 10 trillion Yuan (USD 1590          encourage data sharing.
                                                         billion) by 2030.

                                                                                                                                                     AI special
You can also read