In October 2017 the State Council of the People’s Republic of China (PRC) – the country’s chief administrative authority – designated artificial intelligence a ‘national priority’, stating its ambition to become ‘the world’s premier artificial intelligence innovation centre’ by 2030. The stakes are high since Chinese AI has medical, industrial and transport applications, particularly in autonomous vehicles, and is potentially a major driver of the country’s growth.
Beyond mere expressions of political will, China has undoubted advantages. Chinese consumers are the world’s biggest users of smartphones to pay for goods and services (1), voice recognition software and virtual assistants. The Chinese public have widely accepted that robots plug labour shortages in hotels, hospitals and banks: it’s not uncommon to encounter delivery or hotel reception robots (2). They are also used in construction, mining, and even disaster relief.
With 800 million smartphone users out of a population of 1.41 billion in 2021, China generates a vast amount of digital information, which is widely accessible due to regulations which, while they protect consumer data, are often relaxed in cases of ‘national interest’ (in the field of health, for example). There are special regulatory regimes for biotechnology, bioengineering and biopharmaceuticals to encourage collaboration between companies, researchers and local authorities. And this close association attracts investors: several major biotech breakthroughs have resulted, such as automatic natural-language processing and machine learning to collect and analyse medical data to predict clinical trial outcomes and optimise the design of clinical studies.
China is undisputed champion in artificial intelligence research papers … far surpassing the US in both quantity and quality
China also has a large pool of skilled workers. About 1.4 million engineers qualify annually, six times as many as in the US, at least a third of them in AI (3). As the Japanese business daily Nikkei Asia noted, ‘China is the undisputed champion in artificial intelligence research papers … far surpassing the US in both quantity and quality’ (4). Tencent, Alibaba and Huawei are now among the top ten companies in the world producing this type of output.
This research sometimes produces real-world applications that improve on existing global models. This is true of the mobile video-sharing app TikTok (see ‘Clock ticking for TikTok’, in this issue), but also of the cloud computing division of Alibaba (a leading e-commerce company). Alibaba’s customer service chatbots and personalised teaching tools in education are being used to customise communication at scale with each individual user, according to their needs. Alibaba’s City Brain software system has improved traffic flow so much that Hangzhou (population 6.97 million) has gone from 5th to 57th in the list of the world’s most congested cities (5). The system monitors congestion points and automatically adjusts traffic lights, an optimisation that proved especially useful for ambulances during the Covid crisis as lockdown was in place, unblocking certain parts of the city whilst others were still congested with traffic.
Success stories in cybersecurity
There are many success stories, particularly in cybersecurity, where there are four major players (Bairong, TransInfoTech, Dahua Technology, Hikvision). And SenseTime is poised to capture the largest share of the domestic market for AI technologies using computer vision. This company, which is already dominant in facial recognition security, is now applying its expertise to other sectors. It has developed software that can identify faces, estimate people’s ages and cross-reference them with commercial data, helping retailers ascertain who is likely to buy their products.
The Chinese industry stands out from its international competitors because of its ability to merge digital power and retailing expertise, through the successful integration of online, offline and logistics data in a single value chain. Coupled with AI, this integration has enabled the creation of an ultra-efficient delivery model. For example, a lipstick ordered from JD.com, another e-commerce giant, can be shipped in six minutes by one of the world’s most advanced logistics centres. In May 2022 Baidu, ‘China’s Google’, launched driverless taxis in Beijing, while WeRide has made more than 150,000 such trips in Guangzhou (population 14 million) without a single accident.
Certainly, in the field of generative AI (a branch that uses existing content to generate new content), the US AI chatbot ChatGPT – which can hold intelligent conversations with users – has stolen a march. But the Chinese giants are creating new apps and looking to compete with their own platforms. Tencent’s Different Dimension Me animated image generator, which arrived on global social media platforms in late 2022, can create anime-style images from photographs of faces, and Baidu, which has a ten-billion-parameter model (ERNIE-ViLG), was launched mid-March 2023. Also worth noting is Idea, a research lab headed by renowned computer scientist Harry Shum, which has invented Taiyi, another large-scale text-to-image model.
Despite the favourable landscape and numerous initiatives, science and technology minister Wang Zhigang acknowledged this March that ‘China must wait’ before it will ‘see results’ (6) comparable to ChatGPT. It was also evident that Chinese AI was of little use in tackling the Covid-19 pandemic. It was employed to restrict people’s movements and predict the evolution of the disease, but played little part in developing effective vaccines, unlike at US startup Moderna.
Several factors explain the failure of the state authorities’ forced-march AI initiative. First, although almost a third of the world’s best AI researchers come from China, only a tenth of them actually work there. Most Chinese talent is in the US, to the point that this is ‘a US secret weapon in AI’, according to a study from US think tank MacroPolo (7). In addition to this brain drain, there are discrepancies between the investments promised and actually delivered, which players in China’s IT industry have publicised as a way of lobbying for funds. This was the case with the $16bn announced in 2018 by Tianjin, a major port city in northeast China; it’s unclear whether the funds were ever released and, if so, if they reached the right recipients and were used exclusively for AI.
Hefty fines or suspension
Beijing’s very strict regulation of this sector has also effectively hampered some advances. Ant Group, the financial arm of Alibaba, whose dual listing in Hong Kong and Shanghai was suspended in November 2020 (8), has had its wings clipped. Industry players found to have breached competition rules receive hefty fines. In July 2022, after a lengthy investigation, the Internet regulator fined Didi, ‘China’s Uber’, 8bn yuan ($1.16bn). The two computer game giants Tencent and NetEase, penalised for antitrust violations in 2021, were barred from bringing new games to market until the end of 2022, sending their stock prices plummeting.
Fearing arrest, several tech executives have resigned, including Zhang Yiming, the founder of ByteDance (TikTok’s parent company), in November 2021; many have fled abroad, to Japan or Singapore. That same month, the major online privacy law passed by the National People’s Congress came into effect, and its imprecise parameters put operators who handle consumer data at greater risk. This tightening of state control may demotivate China’s digital giants.
What is certain is that the emphasis on AI development will be maintained: Xi Jinping mentioned technology 40 times in his report to the 20th Communist Party Congress last October, compared with 17 times in 2017
In addition, at the start of this year, the government abruptly suspended its $148bn investment plan for the semiconductor industry, which included several aid programmes to support production and research, mainly through subsidies and tax credits. This came after evidence of corruption was revealed through a wave of investigations into leading figures, including Xiao Yaqing, former industry and information technology minister, and Ding Wenwu, who headed the China Integrated Circuit Industry Investment Fund (CICIIF). Informally known as the ‘Big Fund’ and endowed with $45bn, this has been the government’s official vehicle for managing its huge semiconductor investments since 2014. It has a complex web of interests and its shareholders include the ministry of finance, state lender China Development Bank, powerful monopoly China Tobacco and telecoms giant China Mobile.
The suspension came at a bad time for President Xi Jinping and his ambitions for technological autonomy. Chips are essential to ongoing innovations in 5G, the cloud, the Internet of Things and many rapidly evolving sectors such as defence and space. They are indispensable in everything from smartphones and computers to ballistic missiles and the automotive industry; China is the world’s largest consumer market for integrated circuits.
Yet the semiconductor value chain is made up of a series of interdependent technologies, which gives a strategic advantage to those who control its choke points (9). A breakdown, shortage, fault or embargo affecting a single link in the chain can have major repercussions. Only a handful of companies have the requisite expertise to make some components. Chips are produced in three main stages: design, manufacture and assembly-testing-packaging. Choke points are concentrated in design software and lithography technology (a key manufacturing stage) – two specialised sectors in which, despite its efforts, China lags far behind competitors, all of which are located in countries allied with the US: Japan, South Korea, Taiwan and some European states.
Washington is pursuing an increasingly strict embargo. In early October 2022 the White House banned US companies from exporting to China chip-making equipment, which is essential for high-performance computing and supercomputers (10), as well as chips 14-nanometres and smaller, the key requirement of high-tech industries. De facto, the ban includes non-US companies, further restricting Beijing’s options. President Joe Biden even managed to win over the Netherlands, where ASML, a leading manufacturer of chip-making equipment, is located, and Japan, where Electron and Nikon are based. In January 2023 these three states signed a cooperation agreement endorsing the embargo. The US administration’s increasing restrictions have significantly sharpened Chinese interest in Taiwan, home to the world’s most valuable semiconductor production firm, Taiwan Semiconductor Manufacturing Company (TSMC), which makes half of the world’s most advanced chips.
These tighter sanctions have hit a major, flourishing sector in China: in 2020 the country accounted for 9% of the global semiconductor market ($39.8bn in total annual sales), compared to just 3.8% in 2015. But it’s highly import-dependent for sophisticated chips ($378bn, or 18% of its total imports, including energy). It also produced slower-speed semiconductors used in computers and home devices. Last year, its exports fell dramatically (a 72.8% drop compared to 2021). Some foreign customers, fearing further restriction by US authorities are now choosing to diversify their supply chains. Dell, for example, plans to stop using chips made in China by 2024 and has already asked its suppliers to reduce the quantity of other components produced there (11). Its rival Hewlett Packard has also begun surveying its suppliers to gauge the feasibility of moving production and assembly.
China’s tech giants are also taking steps to protect their international business interests. Some startups have left the country altogether: Movio (formerly Surreal), a generative AI video platform for video marketing, has relocated to Los Angeles to sell its services on the global market, despite its considerable success in China during the pandemic.
For now, the Chinese industry still has significant reserves of semiconductors in private and state laboratories. But longer term, the shortage could threaten its ability to generate its own powerful algorithms, which are vital for developing competitive AI. This could affect companies that use US-sourced hardware, especially for autonomous vehicles and logistics, as well as research centres using AI for drug discovery.
Of course, Beijing has not been passive. The authorities are now asking local suppliers to reduce their component prices for domestic semiconductor customers, both to boost export competitiveness and to reduce dependence on imports, which have hitherto been cheaper. However, Chinese leaders are divided on how to respond to US measures. Some advocate even higher investment to close the gap – there is talk of an additional $145bn plan despite the Big Fund affair, which acted as a wake-up call – others are banking on indirect support for major projects that use Chinese-made chips, such as 5G. What is certain is that the emphasis on AI development will be maintained: Xi Jinping mentioned technology 40 times in his report to the 20th Communist Party Congress last October, compared with 17 times in 2017.
Washington’s aggressive attitude will certainly delay, but probably not derail, these fresh efforts. It may lead Beijing to decouple from the rest of the world and result in divergent technical standards, hampering the global potential for collaboration on new technologies. It will take some time to discern the full effects, but at the very least it’s slowing down Chinese innovation in the short term.
https://mondediplo.com/2023/04/05china-ai China’s race to become global AI superpower, by Gabrielle Chou (Le Monde diplomatique