China's AI plan lays foundation for long-term strength

Analysis

China's national-level AI development plans are aspirational, doing little to establish priorities or allocate resources and clearly define responsibilities (see CHINA: Artificial intelligence could transform China - September 21, 2017).

However, they establish AI overall as a high national priority, explicitly state an ambition to lead the world in the field by 2030 and signal that initiatives in AI by subsidiary levels of the state system will be rewarded.

Decentralised implementation

Concrete policies are left to ministries, local governments and public institutions such as universities and the Chinese Academy of Sciences.

By mid-April this year, 18 provinces and cities had their own AI development plans, with a combined target output of 400 billion renminbi in 2020.

For example, Beijing municipality announced in January a 13.8-billion-renminbi AI development park, and Tianjin announced in May a 100-billion-renminbi fund to subsidise AI research institutions to set up in the city and attract top talent with salaries of up to 2 million renminbi.

The real figures will probably fall short of those announced in many cases but will be very significant nevertheless.

Investment bubble?

Local governments will follow political signals by ploughing resources into AI projects. Inter-regional competition will inflate the costs of incentives to attract businesses and talent as regional centres vie to become 'China's Silicon Valley' (China has no one leading technology hub). Private firms will leap into an area they know enjoys government favour and funding.

The result may be overcapacity and the collapse of many unviable investments (see CHINA: Monopolies and abuses hold back tech sector - February 12, 2018).

A possible warning sign is the seemingly opportunistic entry of firms with no relevant experience. For example, real estate conglomerate Evergrande announced in April a 100-billion-renminbi investment in three cutting-edge technology centres, with AI as one focus.

However, even investments that miss their desired objectives may create capacity that can be recycled for other, more successful innovative activities. This occurred in the United States after the end of the space race, when thousands of laid-off engineers built much of the US tech industry.

Government funding

Enterprises will carry out the majority of R&D, but often in partnership with public institutions or with state funding.

The central government will directly fund some projects. For example, the first half of 2018 saw the Ministry of Science and Technology allocate 2.73 billion renminbi to eight AI-related research projects.

However, 'government guidance funds' could be a much larger source. These are established mainly by provincial and city governments, and function like venture capital funds, raising private capital and adding public money to invest in start-ups in priority sectors. By end-2016, around 1,000 such funds had raised 1.9 trillion renminbi. Their two top priorities now are healthcare and AI. Most recently, Shanghai municipality launched on July 4 a 100-billion-renminbi AI investment fund.

The rapid launch of new funds could contribute to an investment bubble, and to its collapse if too many funds reach maturity simultaneously and M&A or stock market demand falls short of this supply.

Public procurement

China's AI industry will benefit from government procurement, with domestic firms having exclusive access in sensitive areas such as security. For example, SenseTime, which in April raised 600 million dollars to become the world's highest-valued AI start-up, valued at 4.5 billion dollars, derives at least one-third of its business selling real-time facial recognition surveillance systems to Chinese police forces.

The scale of military research and procurement is undisclosed but will be significant. The government promotes 'military-civil fusion', in which the military forges links with civilian research institutes, universities and private firms to pool resources and personnel, develop dual-use technologies, adopt shared standards and avoid duplication.

AI is a particularly attractive field for military-civil fusion because the private sector is on the cutting edge, rather than the state-owned defence firms that traditionally supply the military.

Private sector pivotal

The technology sector is unusual among China's major industries in that large private firms dominate it, namely Baidu, Alibaba and Tencent. These will be the preferred partners and major beneficiaries of public spending on AI. For example, the National Development and Reform Commission will fund Baidu's national deep-learning lab.

China's tech giants will come under heavy political pressure to contribute their resources to state objectives. Two months after the publication of the national AI strategy last year, Baidu launched a 1.5-billion-dollar fund to invest in AI companies. The month after that Alibaba announced a 15-billion-dollar programme to build seven labs in four countries focusing on AI and quantum computing. Such moves are not necessarily taken simply to please powerful politicians; state and commercial objectives will often converge.

These private firms are also crucial to China's AI development because they collect the huge datasets necessary for training machine-learning algorithms. They will come under pressure to share them, for example, through partnerships with universities such as that announced between Alibaba and Tsinghua University in March.

The Ministry of Science and Technology last November designated the three largest private firms, Baidu, Alibaba and Tencent, as 'national champions' in developing self-driving cars, smart cities and computer vision for medical diagnostics respectively, and a fourth firm, iFlytek, the leader in speech recognition. These designations may indicate the central government's priority areas.

The national champions have good prospects. They are already world-class firms and can expect state support in their efforts to compete overseas and play a policy role domestically (such as assisting urbanisation through 'smart cities'). That means a huge captive market.

Technology transfer

The most controversial elements of China's AI policies will be its promotion of inward and outward foreign investment as a means to catch up with the world leader, the United States.

China has had some success in attracting foreign firms. Google announced plans last December to open an AI research centre in Beijing. Microsoft last month revealed plans to cooperate with four Chinese universities to create an open AI platform.

However, China will have to calibrate its efforts carefully. US firms may be deterred if Chinese partners apply too much pressure to share technology. For example, AmCham China reported last year that 36% of US firms held back investment because of technology transfer requirements. These requirements also risk political backlash; they are among the major grievances cited in the 'Section 301' report that underpins the Trump administration's tariffs on Chinese exports.

Overseas acquisitions

Chinese investments in the United States will sometimes face barriers too.

Baidu opened an AI lab in Silicon Valley in 2014 and plans another in the near future. Tencent set up an AI lab in Seattle in May, led by a former Microsoft scientist. Eleven Chinese 'accelerators' in San Francisco provide assistance to technology start-ups, some of them using government funding. Such initiatives give Chinese firms access to talent unavailable in China, including Chinese graduates of US institutions who do not wish to return.

US politicians and officials cite AI as a field with implications for national security when arguing for restrictions on Chinese investments in the US technology sector. Legislation now going through Congress would expand the powers of the Committee on Foreign Investment in the United States (CFIUS), which screens incoming foreign investment.

Fears will deepen when the cyberespionage and conventional espionage that will form an unspoken part of China's AI strategy are occasionally exposed.

Human resources

The Ministry of Education in April issued an 'AI Innovation Action Plan for Colleges and Universities' and a programme to train 500 teachers of AI and 5,000 students at top universities. The previous month it approved 2,311 new majors in high-technology subjects, including data science, robotics and AI. The first school textbook specifically on AI was published in April and pilot classes have been introduced in 40 high schools. A scholarship programme will fund Chinese postgraduate students to study AI in North America.

Training new talent in China will be a slow process because it requires hiring qualified instructors who must either be trained themselves or attracted from overseas.

Chinese recruiters visit US universities to head-hunt experts. The AI plan calls in particular for the recruitment of world-class scientists through the 'Thousand Talents' programme and similar initiatives, which offer scientists grants, higher salaries and more generous perks than they can get overseas.

The decentralisation of foreign talent recruitment makes definitive totals impossible, but state news agency Xinhua last year claimed that 40,000 "high-end professionals" have been recruited through the Thousand Talents programme since its launch in 2008. The US government in April put the current number at 2,629.

Overseas recruitment efforts struggle against the disadvantages of living in China compared to US industry centres such as California. These include pollution, poor choice of schools for children, censorship and conflicting political values or loyalties. Higher salaries cannot always compensate.

China may have more success attracting professionals from the Global South, for whom the pay hike may be relatively greater and the drop in living standards less severe.

The Trump administration's moves to restrict visas for Chinese science and technology students and restrict Chinese nationals from projects deemed sensitive could aid China's efforts to retain and repatriate talent.

Standard-setting

China's government eyes a leading role in setting technical standards for AI, which is still in its early stages.

The Standards Administration of China issued a white paper on AI standardisation in January to coordinate the national effort. In April, Beijing hosted the first ever meeting of 'Subcommittee 42', established under the International Organization for Standardization and the International Electrotechnical Commission to coordinate AI standard-setting internationally.

Setting standards will help the industry by promoting interoperability and allowing the pooling of data. It will also reduce the risk of accidents or scandals that turn public opinion against AI.

Participating in setting global standards provides the opportunity to push for Chinese patents to be incorporated, bringing royalties to Chinese firms, and ensures that global standards are not crafted in a way that disadvantages China's large firms (or the Communist Party).

International exchange

Institutions and local governments in China organise academic and commercial conferences and forums in China. Such events do not always live up to their grandiose names, but do attract international participation. Microsoft attended the 'World Intelligence Congress' hosted by Tianjin in May, and Intel is supporting a competition related to the central-government-backed 'AI World 2018' conference in Shanghai in September.

Chinese firms and institutions also pursue partnerships with Western institutions, such as the five-year research agreement signed between Chinese voice-recognition firm iFlytek and MIT's AI lab in June.

Chinese researchers are encouraged to participate in international conferences. Last month, China's AI Industry Alliance co-hosted a 'US-China AI Tech Summit' with the AI Alliance of Silicon Valley and the Future Society, attended by academic researchers and representatives from major US and Chinese tech firms.

Such exchange is likely to be well received in the West, not just because of the advanced technical content China contributes, but because some researchers fear that a country or group that reaches a major AI breakthrough might deploy it without adequately considering the ethical and safety implications (see INTERNATIONAL: The singularity is distant - October 11, 2017 and see INTERNATIONAL: Regulating AI - December 5, 2017). This concern is greater vis-a-vis states with opaque and authoritarian political systems such as China's.