In July 2017, China’s government published an ambitious policy paper, outlining how the country would become the world leader in AI by the year 2030. But by some measures China has already succeeded in this goal — a decade ahead of schedule.
A new study shows that China’s output of influential AI research papers will soon overtake that of the US, the world’s current number one in AI research. The finding suggests that China’s plan to expand its AI capabilities with the help of generous government investment in both educational facilities and private industry is paying off.
In terms of sheer volume of AI papers published each year, China surpassed America back in 2006, but critics have pointed out that quantity does not necessarily equal quality. China’ has well-documented problems with scientific fraud, and in AI there is a stereotype of Chinese research as incremental. For these reasons, some have suggested that counting the sheer number of papers is not necessarily a meaningful metric for AI achievement.
But new research from the Seattle-based Allen Institute for Artificial Intelligence (AI2) accounts for this by measuring not just the number of papers, but how often they are cited, a good shorthand measure for influence in the wider community.
After analyzing more than two million AI papers published up until the end of 2018, the Allen Institute found that China is “poised to overtake the US in the most-cited 50 percent of papers this year, in the most-cited 10 percent of papers next year, and in the 1 percent of most-cited papers by 2025.”
The researchers found that America’s share of the most-cited 10 percent of papers declined from a high of 47 percent in 1982 to a low of 29 percent in 2018. China’s share, meanwhile, has been “rising steeply,” reaching a high of 26.5 percent last year.
Oren Etzioni, a professor of computer science and CEO of the Allen Institute, told The Verge that this research “refutes” the stereotype of Chinese contribution to AI as incremental.
“Clearly, the quality of Chinese papers is high and getting higher,” said Etzioni over email. “Of course, one might argue that citation statistics could be influenced by more Chinese scientists citing each other, but if you look at the list of Best Paper Awards you’ll see several Chinese entries there which represents the absolute cream of the crop.”
The Institute’s does note, though, that when it comes to the number of Best Paper awards in computer science — a somewhat “idiosyncratic” measure which depends on the ebb and flow of trends and interests in various fields — the US is still “firmly ahead.”
These findings should make for interesting reading for the US government. Although analyzing research is only a single metric when it comes to measuring the overall AI output of any country, academics and industry experts have warned for years that America needs to do more to maintain its lead.
Last month, President Trump signed an executive order intended to spur investment in AI, but the order was vague, with few concrete goals, and included zero new funding for research. Other countries’ national AI strategies have included government investment ranging from $20 million (in Australia and Denmark) to nearly $2 billion (in South Korea).
More importantly, Trump’s initiative failed to address what many see as the biggest challenge in the field: attracting global talent. The number of top-flight AI researchers is limited, and the current US approach to immigration is the opposite of what’s needed.
As these new figures from the Allen Institute show, the development of cutting-edge AI is now very much a global affair, and isolationism will not help anyone. “We need more funding,” says Etzioni, “and even more importantly, the administration has been discouraging talented students from coming and staying in the US.”