Advanced AI is currently pretty much a duopoly between the USA and China. The US is the clear leader, thanks largely to its tech giants – Google, Meta, Microsoft, Amazon, and Apple. China also has a fistful of tech giants – Baidu, Alibaba, and Tencent are the ones usually listed, but the Chinese government has also taken a strong interest in AI since Deep Mind’s Alpha Go system beat the world’s best Go player in 2016. Are the world’s two AI superpowers doomed to confrontation and conflict?
Western misperceptions of China
People in the West don’t know enough about China’s current and future role in AI. Some think its companies just copy their Western counterparts, while others think it is an implacable and increasingly dangerous enemy, run by a dictator who cares nothing for his people. Both these views are wrong.
One person who has been trying to provide a more accurate picture of China and AI in recent years is Jeffrey Ding, author of the influential newsletter ChinAI. Ding is currently an Assistant Professor of Political Science at George Washington University, having earned a PhD at Oxford University, and spent some time with the Future of Humanity Institute there.
The Thucydides Trap
According to a theory in international relations known as the Thucydides Trap, the mere existence of a rising new power is likely to lead to conflict with an established power. It is named after a Greek historian who suggested that the rise of Athens meant that war with Sparta was inevitable. We must all hope that America and China can avoid falling into this trap, as the consequences would be disastrous.
It is hard to measure the extent of any country’s AI capabilities because organisations often don’t report it separately. The value created by an AI system optimising the routing of a company’s drivers, for instance, or monitoring the quality of goods on a production line, would be reported as part of the country’s manufacturing industry, not as part of its AI industry.
It is also hard to measure the level of deployment of AI because it is still fairly low in all countries, including the US and China. Deep learning, the type of AI which came to prominence in the Big Bang of 2012, is still fairly new, and it is still difficult and expensive to deploy. But what can be measured is the deployment of enabling technologies like computer hardware and cloud computing. Using these as proxies, the US is well ahead of China.
More forward-looking metrics include the number of papers published by each country, the number published in prestigious journals, and the number of patents granted for AI applications. On these metrics, China is still some way behind the US, but the US and China are a long way ahead of everybody else.
EU countries are laggards in AI development, but perhaps not in deployment
In particular, it is remarkable how far behind European countries are, both individually, and collectively in the form of the European Union. The EU is very active in AI regulation, but it is not active at scale in AI development. However, we are still in the early stages of AI development and deployment, so there is time to rectify this.
Ding thinks that one of the biggest misapprehensions in the West about AI in China is the belief that China is further ahead than it really is. He thinks that China is a clear leader in terms of developing advanced AI, but may not be far ahead of European companies in the deployment of AI.
The Chinese do care about privacy
Another common misperception in the West is that the Chinese government is able to use its citizens’ data without restraint, because Chinese people just don’t care about privacy in the way that their counterparts in the West do. In fact, surveys show considerable concern in China about the use of face recognition systems, which have been deployed widely there.
On the other hand, the widespread perception that China’s leaders at both national and regional levels are more focused on the opportunities of AI than leaders in the West is accurate. The July 2017 National AI Plan has been echoed at all levels of administration, and implementation has been relatively consistent.
Of course, just because a government spends a lot of money promoting an industry does not mean it will necessarily succeed. The Japanese government’s fifth generation computing initiative was widely expected to give the country a major lead in advanced technologies – including AI – but it turned out to be backing the wrong technological horses.
Social Credit
Associated with the notion that the Chinese government uses its citizens’ data recklessly and without restraint, is the idea that the country’s Social Credit system is a unitary structure which surveils the people relentlessly, and allows the government to exert a level of control that would make Orwell’s Big Brother envious. In fact, Social Credit is the name for a collection of loosely integrated systems which are often more akin to the credit scoring services familiar in the West than to the stuff of Orwell’s nightmares. Social Credit does include regional blacklists under its banner, but these are not integrated at the national level, and they do not seem to be using AI or other sophisticated technologies.
When the system was first conceived in 2014, a more comprehensive system for nudging people toward good behaviour was envisaged by some officials, but it turned out to be hard to assemble the most draconian versions imagined. Western perceptions often veered to the more extreme possibilities regardless.
De-coupling at different speeds
At the international political level, there are clearly grave tensions between the US and China. President Trump liked to rouse his supporters by imposing sanctions on Chinese companies, and threatening grave consequences for perceived Chinese transgressions. President Biden has maintained a sceptical attitude, and has tightened restrictions of the supply of advanced technologies to Chinese organisations.
But Ding thinks that with the exception of specific technologies like advanced semiconductors, collaboration and co-operation is alive and well between companies and between universities, and he thinks this is actually likely to deepen. The same may be true of finance, where the growth of Chinese companies has been enabled to a significant degree by Western venture capital funds and other sources. For instance, Alibaba is very much a Chinese company, but until recently, Yahoo owned 40% of its equity.
This financial globalisation is probably easier to unwind than the industrial and academic globalisation. A number of Chinese companies have de-listed from Western stock markets and retreated to the exchanges in Hong Kong and Shanghai. Financial de-coupling is driven as much by the Chinese government’s desire to ensure that Chinese entrepreneurs can never threaten the primacy of the Party as by any initiatives taken by Western governments. In 2020, Alibaba’s proposed IPO of its financial services arm Ant Financial was halted by the government, in the start of a crackdown on tech firms. Post-Covid, the government is keen to rekindle economic growth, and this crackdown seems to have been lifted.
It’s good to talk
However these financial, commercial, and academic relationships evolve, what is of paramount importance is that the leaders of the world’s two most powerful countries continue to talk to each other openly and constructively. The worst possible outcome for the world is war between them. To coin a phrase, it’s good to talk.