Kai-Fu Lee, a prominent investor and entrepreneur based in Beijing, has been talking up China’s artificial intelligence potential for a while. Now he’s got a message for the United States. The real threat to American preeminence in AI isn’t China’s rise, he says—it’s the US government’s complacency.
Lee is well placed to understand the issue, even if he isn’t altogether unbiased. He worked on machine learning at Carnegie Mellon University during the 1980s, led Microsoft’s research lab in China in the 1990s, and then spearheaded Google’s venture into China in the 2000s. Today Lee leads Sinovation Ventures, an AI-focused incubator based in Beijing. He is also the author of AI Superpowers, a new book that explores the Chinese and American AI booms.
Rather than competition from China, Lee says, the real risk for the US is in failing to invest in and prioritize fundamental AI research—a problem that’s being exacerbated as big US companies suck up much of the top talent in the field. In general, tech firms focus less on fundamental breakthroughs than does academia, which struggles to compete with the private sector in retaining researchers.
“The US should set out some really big challenges that the current technology cannot solve,” Lee told MIT Technology Review. A resurgence of interest in AI has been inspired, in large part, by stunning advances in deep learning, a technique that uses very large artificial neural networks to learn from data. But the approach requires huge quantities of data, and it tends to work only in narrow domains. “A next set of technologies [is needed] to overcome the limits of deep learning. Commercial companies aren’t going to focus on these things.”
It is, perhaps, easy to see AI as a race that China is winning. The country’s vibrant tech industry has already adopted AI at a remarkable rate. Last year the country’s government also announced a sweeping plan to advance its AI industry (see “China’s AI awakening”). The US government, meanwhile, has taken a hands-off approach, relying on its robust tech industry to power developments in AI (see “Here’s how the US needs to prepare for the age of artificial intelligence”).
China’s scale, entrepreneurial zeal, and central planning are putting the country in a better place than the US to commercialize AI going forward, Lee says. But he says the US still has a big advantage in fundamental research, which it shouldn’t squander. The Trump administration should follow the lead of China as well as France and Canada by investing heavily in AI research. “Doubling funding is not at all outrageous,” he says.
Lee also has a creative solution to the AI brain drain in academia. Given that industry researchers can earn several times as much as their academic counterparts, AI professors could be subsidized so that they earn twice as much as they do now, he says, and the government could build big-time computing resources to help support their work.
“Can there be a gigantic GPU farm that gives professors as much computing power as Google has?” Lee says, referring to data centers filled with the graphics chips that are used to train deep-learning models. “Can you use the National Institute of Standards and Technologies to create a gigantic database that will neutralize the advantages of Facebook?”
The Trump administration’s immigration policy is another liability, Lee says. The US has benefited enormously from bringing the best international students and professors to its academic institutions, so curtailing them seems foolish. And the idea that Chinese students are in the US to steal technology—a concern often cited in policies that restrict Chinese students’ access to visas—is offensive, he argues. “There’s no basis to make that kind of accusation,” he says. “They are normal students.”
Not long ago, Lee made a name for himself as a kind of hype man for China’s AI scene. That helped bolster the reputation of Chinese AI tech companies—including many in which his firm, Sinovation Ventures, had a vested interest. These days, his argument that the US and China should work together on developing AI also happens to be the same stance that’s being softly pushed by executives and politicians in China, including President Xi Jinping (see “China’s leaders are softening their stance on AI”).
Biased though they may be, his assertions make a certain amount of sense. In stressing that both countries can succeed at the same time, he points to the fact that Chinese and Western companies rarely compete in each other’s markets. “We are largely operating in parallel universes, both of which can succeed and grow,” Lee says. “It absolutely isn’t a zero-sum game.”
Lee sees both China and the US flourishing outside their home markets. In fact, he believes AI could usher in a new world order largely defined by these two technology superpowers. “We are kind of on the path to splitting the world in half and expanding these two parallel universes beyond the countries themselves,” Lee says.
Behind the boom
As each country’s AI giants expand aggressively around the world, US firms—Google, Amazon, Microsoft, Facebook, and the like—are focusing more on markets in developed countries, while China’s champions—Alibaba, Tencent, Baidu, and upstarts like SenseTime—have their eyes on developing countries.
This pattern of expansion is likely to define competition between the two countries over the next decade or so. But its long-term effects aren’t yet well understood. As firms gather more data, build more-powerful computing resources, and hire more experts, they will likely become ever more difficult to challenge. This will mean that the technology supplied to institutions around the world will increasingly be conceived and defined by US and Chinese tech culture. The face-recognition services offered by both Chinese and US companies, for example, could become a central aspect of policing in other parts of the world. This could lead to a new kind of techno-cultural colonization.
Lee is cautious not to pick winners in the race to run the world with AI. “I don’t think China will necessarily dominate,” he says. But he suggests that China might be more welcome than the US in many places. “Having been technically colonized by the West, China has empathy to help other countries develop their own industries and identity,” he says.
Artificial intelligence is creating a new colonial world order
An MIT Technology Review series investigates how AI is enriching a powerful few by dispossessing communities that have been dispossessed before.
Meta has built a massive new language AI—and it’s giving it away for free
Facebook’s parent company is inviting researchers to pore over and pick apart the flaws in its version of GPT-3
This horse-riding astronaut is a milestone in AI’s journey to make sense of the world
OpenAI’s latest picture-making AI is amazing—but raises questions about what we mean by intelligence.
How the AI industry profits from catastrophe
As the demand for data labeling exploded, an economic catastrophe turned Venezuela into ground zero for a new model of labor exploitation.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.