China battles the US in the artificial intelligence arms race

What counts is implementation not innovation, and here the Chinese have big advantages

Martin Wolf

In late March I attended the China Development Forum for the ninth time. The visit stimulated my recent observations on China’s economy and politics. But what makes the CDF most valuable is serendipity. This time that came in the shape of a meeting with Kai-Fu Lee, former president of Google China and now a leading venture capitalist in Chinese technology.

Mr Lee gave me a copy of his new book, AI Superpowers: China, Silicon Valley and the New World Order. This has a startling story to tell: for the first time since the industrial revolution, he argues, China will be at the forefront of a huge economic transformation — the revolution in artificial intelligence.

He starts his book by talking about China’s “Sputnik moment”, when Google DeepMind’s AlphaGo defeated Ke Jie, the world’s leading player of the ancient Chinese game of Go. This demonstrated the capacity of modern AI. But, by implication, Mr Lee’s book foresees another such moment, when the US realises it is no longer leader in the global application of AI. The original Sputnik moment occurred when the Soviet Union put the first satellite in orbit in 1957. This led to the space race of the 1960s, which the US duly won. What will the present “race” lead to?

Mr Lee does not claim that China will lead in fundamental innovation in this area. But that may not matter, since the big intellectual breakthroughs have already occurred. What matters most is implementation, not innovation. Here China has, he writes, many advantages.

First, the work of leading AI researchers is readily available online. The internet is, after all, a superlative engine for spreading intellectual breakthroughs, not least including those in AI.

Second, China’s hypercompetitive and entrepreneurial economy lives by Facebook founder Mark Zuckerberg’s notorious motto: “move fast and break things”. Mr Lee describes a world of cut-throat business activity and remorseless imitation, which has already allowed Chinese businesses to defeat leading western rivals in their home market. The ceaseless “trial and error” of the Chinese business model is, he argues, well suited to rolling out the fruits of AI across the economy. It could, for example, work far better in introducing autonomous vehicles than the west’s safety-conscious approach. China’s swarms may be inefficient, but they are effective. That is what matters.

Third, China’s dense urban settlements have created a huge demand for delivery and other services. “American start-ups like to stick to what they know: building clean digital platforms that facilitate information exchanges,” Mr Lee argues. But Chinese firms get their hands dirty in the real world. They integrate the online and offline worlds.

Fourth, China’s backwardness allowed businesses to leapfrog existing services. So China has been able to jump to universal digital payment systems, while western businesses still use outdated technology.

Fifth, China has scale. It has more internet users than the US and Europe combined. If data is indeed the fuel of the AI revolution, China simply has more of it than anybody else.Sixth, China has a supportive government. Mr Lee cites a speech by premier Li Keqiang in 2014 at the World Economic Forum’s “summer Davos”, calling for “mass entrepreneurship and mass innovation”. In his report “Deciphering China’s AI Dream”, Oxford university’s Jeffrey Ding points to the State Council’s national strategy for AI development. China’s government has ambitious goals and is willing to take risks to achieve them. One of the things China can do more easily than anywhere else is build complementary infrastructure.

Finally, writes Mr Lee, the Chinese public is far more relaxed about privacy than westerners.

Chinese leaders, it may be argued, see no justification for individual privacy at all (except for their own).
 So where is this supposed “race” between the US and China today? Mr Lee distinguishes four aspects of AI: “internet AI” — the AI that tracks what you do on the internet; “business AI” — the AI that allows businesses to exploit their data better; “perception AI” — the AI that sees the world around it; and “autonomous AI” — the AI that interacts with us in the real world. At present, he thinks China is equal to the US in the first, vastly behind in the second, a little ahead in the third, and, again, far behind in the fourth. But five years from now, he thinks, China might be a little ahead in the first, less far behind in the second, well ahead in the third and equal in the last. There are, to his mind, no other competitors.

Mr Ding analyses the drivers differently. He distinguishes hardware, data, research and the commercial ecosystem. China is far behind the US in production of semiconductors, ahead in the number of potential users and has about half the number of AI experts and roughly half the number of AI companies. All told, China’s potential is about half that of the US. Yet Mr Ding is looking at AI overall, while Mr Lee focuses on commercial applications.

Historical experience suggests that the rents created by a lead in an important technology are valuable, though often impermanent. So, which country will be ahead in the application of AI is indeed important. But the economic and social impact of AI is a bigger issue and one that is relevant to every country.

As Mr Lee stresses, advances in AI offer gains. This is not just in personal convenience, but in improving medical diagnostics, tailoring education to individual students, managing energy and transport systems, making courts fairer, and so on and so forth.

Yet AI also threatens huge upheavals, notably in labour markets. Many of the jobs (or tasks) that AI might do are today done by relatively educated people. It seems reasonable to fear that AI will accelerate the hollowing out of the middle of the earnings distribution, possibly even the upper middle, while increasing concentrations of private wealth and power at the top.

Yet perhaps the most important consequence will be in the intensity of influence and surveillance made possible by AI-monitored mobile devices and sensors. George Orwell’s Big Brother (or many big commercial brothers) might watch us all the time. Such perfect monitoring might be attractive to China’s state. It is horrible to me and, I hope, billions of others.

AI, Mr Lee insists, is not the same as artificial general intelligence: the true super brain is far away. Even so, the challenges this AI creates are huge. We will not stop it. But we may in the end conclude we have birthed a monster.

0 comentarios:

Publicar un comentario