How China's Artificial Intelligence Strategy Stacks Up Against U.S. Investments

Artificial intelligence is becoming a key factor in the future of economic and military competition around the world.

China has been a leader in the field, with heavy investments in AI research and development, and President Trump wants to help the U.S. play catch-up to China with a new AI strategy. Trump signed an executive order this month that gets “the ball rolling on increasing investment in artificial intelligence," says Gregory Allen (@Gregory_C_Allen) with the Center for a New American Security.

“The actual budget totals of what the federal government is going to spend on AI, we don't really know yet,” he tells Here & Now's Eric Westervelt (@Ericnpr). “I do think it is safe to say though that historically over the past years, China has demonstrated significantly greater levels of strategic focus and funding when it comes to artificial intelligence."

For example, Allen says, the Defense Advanced Research Projects Agency is spending the most on research and development at $2 billion over five years. In China, the province of Shanghai, which is a city government, is planning to spend $15 billion over 10 years.

“So literally we have the U.S. federal government at present at risk of being outspent by a provincial government of China,” Allen says. “They are really spending jaw-dropping sums on this area.”

While the U.S. government is doing important work in AI research, Allen says it has been so far “heavily outspent” by private companies in Silicon Valley. Whereas in China, AI research and development is a priority for the government.

“[Chinese officials] say rather openly their goal for artificial intelligence is to transform their military and make it into the strongest military in the world,” Allen says. “Their hope is that they will have an advantage in adopting AI technology, and they will use that to build a military that is stronger than that of the United States.”

On the meaning of artificial intelligence

“The term artificial intelligence, what it means, has changed over time. What we currently use today as tax preparation software such as TurboTax was very successfully marketed as artificial intelligence in the 1980s, but nobody would call that AI today. The most recent innovations around artificial intelligence relate to machine learning, which is a sort of new domain of computer programming whereby rather than a human typing in every line of code by hand, the machine — based on applying an algorithm to a data set — in a sense programs itself. And that's a bit of oversimplification, but it really does get at the heart of what machine learning is and also what kind of capabilities that it enables.

“So many things that in the past we would have said are impossible for computers to automate are suddenly on the horizon. There was actually a major book in 2005, the essential gist of the book was what computers can't do, and making a left turn into ongoing traffic was a canonical example of something that is impossible to automate. And that was true before machine learning really hit its stride in the past five or seven years. Nobody would say that driverless cars are impossible, and in fact, they appear to be right on the horizon.”

"It really does go all the way up to President Xi Jinping in China who really has said AI is a priority for this government all the way down."

Gregory Allen

On how machine learning can be applied to various types of technology

“I mean, machine learning is why language translation has become so high quality of late, why voice recognition software, which used to be awful, is suddenly quite good. It's at the heart of the Amazon Echo — the ability to hear your voice, translate that into actual text that can be used for a computer interface. The same is also true with image recognition. It used to be quite difficult for a computer when given an image to say what is in the image. Now that's a really high performance capability, and in fact in some areas, you can actually beat human experts using AI systems.”

On China’s strategy in aggressively investing in AI

“I went to China four times in the second half of 2018 and had an exciting opportunity to engage with Chinese diplomatic officials, military officials, an executive in China's technology sector. I'll say the first thing that struck me was the level of strategic prioritization of AI. It really does go all the way up to President Xi Jinping in China who really has said AI is a priority for this government all the way down.

“And the other thing that really struck me was the strength of Chinese companies in this area. Just one example, using machine learning for voice recognition — the achievement of a system that could beat average human performance using an AI system — that occurred in China actually before it occurred in the United States. So they really are making real progress in advancing the state of the art in AI research and development. They're also having tremendous success at commercializing their AI research into companies and venture capital ecosystems. All of that is quite strong in China.

“The other thing that I thought was quite interesting was the connection of all of this effort in China to military. The official Chinese AI strategy declares a policy of what translates into English as, 'military civil fusion,' whereby commercial AI developments are then leveraged into military technology. That's a real contrast from the United States. The Department of Defense has rather openly said it's trying to engage with the innovators in Silicon Valley and having difficulties doing so. In China, there are a lot more tools to either legally coerce the cooperation of commercial companies or just to incentivize them with lots and lots of money.”

On China’s weaknesses in AI

“China is extremely strong in research and development. If you go to any international AI conference, Chinese researchers will be presenting their papers there. They will often be quite good papers. They're also strong at the venture capital ecosystem and turning research into commercial companies. But they are not equally strong in all areas. There are areas that they identify correctly as their own weaknesses.

“So the first one I would point out is that of top talent. A Chinese university, Tsinghua University, did a very interesting sort of global census of AI talent, and what they found is that while the United States and China had the largest AI talent ecosystems and are roughly equal there, when it comes to top talent — the sort of the best of the best in the AI research and development community — China actually ranks ninth rather than No. 2. They actually identify themselves as having fewer than 1,000 individuals in the entire country of 1.3 billion individuals who really rank in the upper echelon of AI talent. It's not entirely clear what level of a disadvantage that actually poses. It has not prevented China from its current level of success in advancing the state of art, in [research and development] and also commercializing. But it is certainly something that Chinese government officials are paying attention to, and they have radically increased their spending in the Ministry of Education to try and develop new training programs and university education programs that put AI first.”

On U.S. policy that could help the U.S. advance in AI  

“I think there's so many precursors that affect your ability to effectively implement AI. You have to be thinking about what types of data are you collecting? How are you storing it? Is it ready to be machine readable? And so I think as the U.S. government is thinking about effective policies for AI, they need to think not only about promoting research and development and advancing the state of the art — although that is important — they also need to think about the policies within government for how we're going to use AI and what can make it easier to adopt AI in the service of executing the government mission. I think that's an important area and one that I hope will be prioritized in the coming months.”

Ashley Bailey produced and edited this interview for broadcast with Todd Mundt. Samantha Raphelson adapted it for the web.