The new “nuclear race” is the race to reach a critical A.I. level, also called the technological singularity, which is likely to secure world supremacy to the power who reaches it first.
During the 19th century, Europe opened a new era for humanity by leading the Industrial Revolution. However, since the second half of the 20th century, Europe has become complacent in a 2nd seat of the Digital & Internet Revolution. Now, when a third revolution has begun, that of AI, Europe finds itself left on the 3rd seat, way behind the US and China.
Developing Artificial Intelligence (AI) is crucial for improving our productivity, competitiveness and, ultimately, for preserving our security and values unaltered. It is no surprise that the United States are the most advanced power in digital technologies. However, the EU and the rest of the world have deeply underestimated China, which has overpassed EU and aims to become the world leader.
In this report we review the state of AI in China (compared to the US and the EU), as regards to policy, research, the magnitude of the commercial ecosystem and related investments abroad.
Discussions about AI in Europe are just at the beginning. There are several approaches and policy documents available, such as the Declaration of Cooperation on Artificial Intelligence, signed on 10th April this year. There are other notable initiatives, including the newly (mid-2018) established European Artificial Intelligence Alliance, a group of experts responsible to research and make recommendations on AI policies. In terms of funding, the Commission has decided this year to increase its annual investments in AI, under the framework of Horizon 2020, to reach a value of 1.5 billion euros during 2018-2020.
However, the policy documents are still unclear, non-binding and thoroughly, not very relevant. High-level discussions about AI are almost inexistent. EC’s level of funding allocated to transform EU into an AI competitor on the global stage probably causes some laughs and gigs in Washington and Beijing. The truth is, AI is not a priority for European leaders, for national governments or for the public opinion. There are several individual nations, such as The United Kingdom, Germany and France that have committed strategies and funds for the development of AI. However, they are incomparably insufficient to world competition. Moreover, it is not only investments lagging behind, but also the lack of congruence between member states in regards to their scientific projects and legislation.
This is part of a wider trend across all technological sectors, EU falling behind in terms of global expenditure on research and development (GERD). US and China surpass EU as a whole both in terms of nominal value and of GDP percentage allocated to research. And gaps are increasing.
In fact, it is not only US and China, but most developed countries around the world allocate a larger percentage of their GDP to research than EU.
Another concerning issue is the fact that in both US and, counter-intuitively, China, research is more market-driven than in Europe. In EU, government funding accounts for a 31.74%, whereas in US for 25.50% and in China, highly-surprising, just 21.26%. This suggests that US and China are better able to carry commercially useful research, capitalizing results into lucrative products and ultimately, into productive gains.
Globally, external investment in AI technologies (VC, PE, and M&A activity) had a compound growth between 2013 and 2016 of almost 40% per year. Investment is highly concentrated in geographic hubs, primarily in the US (San Francisco Bay Area, Silicon Valley, Boston and New York), secondly – China (Beijing and Shenzhen) and thirdly – Europe (London). In 2016, the United States absorbed 66 percent of external investment, China was a distant second with 17 percent and Europe further behind.
China’s AI dream
With its biggest tech companies driving momentum in AI research and applications, recently China secured a strong position as the world’s second largest hub for AI development. According to official policy documents, the country aims to become the world leader in AI by 2030 with a related market of 150 billion $. However, it still has a long way to improve in order to be an authentic competitor to the United States.
While the nation’s population ages and economic growth is slowing, China needs to find innovative ways to accelerate productivity growth. AI benefits from unparalleled government support and a favourable regulatory environment. The vast population and diverse industry mix provide a massive volume of data and an enormous available market. However, China is also threatened by the potential disruptions on the labour market that AI can trigger. It is estimated that 51% of work activities in China could be automated – more than any other country in the world.
AI is already a crucial instrument for governance, as the State Council designates AI with an “irreplaceable” role in maintaining social stability and public security. Local governments deploy related technology in other public services such as medical care and judicial activities. Nevertheless, China is actively deploying AI-based social credit systems and sophisticated surveillance mechanisms.
AI is one of the new focal points in policy planning, first emphasized in the Internet Plus initiative launched in 2015. Released in 2016, The 13th Five-Year Plan for Developing National Strategic and Emerging Industries, Made in China 2025 and Robotics Industry Development Plan(2016-2020), followed to provide guidelines for R&D in the field of AI. In February 2017, Artificial Intelligence 2.0 was designated as a mega-project along with fifteen other core technologies. In July 2017, The State Council released the Next Generation Artificial Intelligence Development Plan, a road map that aims to make China an AI world leader and reach a related industrial gross-output worth of $150 billion by 2030.
In November 2017, the Ministry of Science and Technology has designated Baidu, Alibaba Group, Tencent Holdings and iFlyTek as champions of artificial intelligence, each being assigned to lead one AI-related industry: self-driving cars, smart cities, computer optics for medical diagnosis and voice intelligence, respectively.
The nation aims to develop indigenous semiconductors and remove its dependency on foreign countries. In October 2017, The Ministry of Science and Technology prioritized chips for artificial neural networks as one of the 13 “transformative” technology projects with a delivery date of 2021. The Government also funds and encourages domestic companies to invest overseas in companies that own chip technology. Domestic companies such as Baidu and Cambricorn already have positive results in designing chips specifically for AI algorithms. However, despite some optimistic signs, overall China’s semiconductor industry has not yet achieved significant breakthroughs and appears to be a bottleneck to indigenous AI. As of 2015, China accounted for just 4% of the global semiconductor production, while the U.S. accounted for 50%.
In the field of supercomputers, China has strenuously invested and has recently surpassed the US as the country with the biggest number of supercomputing facilities in the world.
Finally, the state is concerned with raising the number and quality of AI talented scholars through a variety of policies and programs, such as the Thousand Talents Program.
To the better or to the worse, the Chinese Government is pro-tech, pro-experimentation, and pro-speed. It provides new technologies with the benefit of doubt, initially encouraging them to grow freely without any restriction, and only in a later phase discussing regulations and addressing other issues. It is not only the central government but also local authorities such as Shenzhen, Shanghai, Nanjing, Wuhan and Tianjin that have significant initiatives for AI development.
If counting the indexed journal articles mentioning “deep learning” or “deep neural network”, China has surpassed the US in both the total number of publications and the number of publications receiving at least one citation. China’s output of academic papers on artificial intelligence overtakes EU as a whole.
However, China lags behind in terms of fundamental research and influence of AI papers.
One of the main reasons is the talent shortage. Despite the largest number of STEM graduates in the world, China has a talent pool of just 39.000 AI researchers, two times less than US.Besides superior numbers, US researchers benefit from more extensive experience, as both the academia and commercial AI ecosystem are more mature. Google alone employs perhaps 50% of the world’s top 100 AI scientists. However, the number and quality of Chinse researchers is fast increasing.
Commercial AI ecosystem
China’s massive economic scale and wide range of industries, an Internet population of 731 million users and rather lax privacy regulations provide AI systems with huge amounts of data and a uniquely fertile environment.
It is hard to estimate exactly the size of China’s AI sector, but it is clear that it had increased significantly in both absolute and relative terms in the past few years. In 2017 China was hosting 23% of world’s total number of AI companies, ranking second after US with 42%.
Every few years, China earns a new entry in Top 10 world’s largest internet companies. In Top 20 there is no European company.
According to a report by iiMedia Research, in 2016 Chinese AI industry had a gross output of 10 billion RMB and was expected to reach 15 billion in 2017. If the Government’s plans succeed and the market reaches RMB 150 billion by 2020, then the AI industry is going to increase tenfold during 2018-2020.
Since the first AI investments in AI in 1999 until 31st June 2017, U.S. companies led with 51.10% (USD 14.8 billion) of world’s total AI funding, while Chinese companies received 33.18% (USD 9.6 billion). Rest of the world, including Europe, account for just 15%.
1999 – 2017 China-US AI industry financing comparison
Between 2014 and 2016, AI industry’s pace of growth in China increased significantly. This period accounts for 55% of all Chinese AI companies ever established and 90% of the total amount ever invested. However in 2017, due to a slowdown of the industry in China, combined with several huge deals in the US, the latter’s share in world AI funding sky-rocketed and U.S. ecosystem nurtured more competitive AI start-ups than China. McKinsey considers the total funds raised by AI start-ups from CB Insight’s AI 100 list, and finds 39 promising start-ups located in the US, while just 3 in China.
Chinese most competitive AI companies include Didi (ride sharing); Meituan (food delivery); Toutiao (news aggregator); Meitu (selfie beautification); Kuaishou (live streaming); Qudian, PPdai and Smart Finance(all fin-tech); 4th Paradigm (solutions to banks and financial institutions); Yibot (customer service solutions for companies such as China Mobile, Didi, Ctrip); iCarbonX (health tracking); Face++ and Sensetime (face recognition); Mobvoi (voice assistant); Ainemo and Rokid (smart home assistants); iFlyTek (China’s leading speech recognition company with a market cap more than double its main US competitor Nuance).
Nowadays, many Chinese firms who have achieved strong success on domestic market are looking to expand globally. Strong financial resources enable them to invest in developed economies, securing brand names, market access, global talents and new technologies. A large part of investments abroad falls in the category of M&As. Technology oriented investments have become predominant in Chinese outbound FDI flows especially since 2017, when the Chinese Government imposed restrictions to other categories of ventures.
United States is the first destination for Chinese outbound foreign investment, and ever since 2014, overall annual Chinese investment in US exceeds US investments in China. As far as the internet tech industry is concerned, the share number of acquisition deals have summited in 2015, but declined in 2016 and 2017 due to the capital controls installed by the Chinese government, combined with higher scrutiny on the American side. The most active Chinese company in US-tech deals is Tencent.
Except M&As, Chinese companies are becoming more present with their own research facilities abroad, in innovation clusters such as Silicon Valley. For example, Baidu operates two AI and Deep Learning research centers in Sunnyvale, CA, whereas Tencent owns a data centre in Silicon Valley and an AI Research facility in Seattle, WA.
Questions about what AI is, what it can already do and what potential it has, cut across technology, economics, politics, ethics and law. The potential of added value is tremendous, as well as the challenges that AI pose. It is certain, however, that AI technologies will continue to make quick breakthroughs that impact fundamentally our societies.
China disposes of huge volumes of data, government support, funds, talent and appetite for technology. Chinese internet giants companies are more and more able to compete with US counterparts, but are still relatively behind in terms of scientific expertise, funds and global-reach. On the opposite side, however, there is no major European company in the field.
It is important to note that China and US should not to be viewed as engaged in a zero-sum competition. Despite the political rhetoric, there is more cross-pollination between the two than one might expect. China is a major market for American hardware, data is shared across borders and researchers from both sides co-author papers. Cross-border AI investments have increased hugely during the past few years. Finally, it is often forgotten the fact that both Tencent and Alibaba are multinational, public companies, owned in significant portions by international stakeholders (Naspers has a 33.3% stake in Tencent and Yahoo has a 15 percent stake in Alibaba).
We are experiencing a technological revolution and the future belongs to those that are the best in seizing AI’s opportunities. At present, European Union is almost totally outside the AI global picture. China and US place a strategic bet on AI and they will both harvest significant results. This paper is meant to be a warning to European stakeholders.
Europe is late, but not too late.
By Adrian Bazavan (PhD candidate in artificial intelligence, based in Beijing)