AI Research: US vs. China

In our AIWS Weekly newsletter last week, we discussed DARPA’s efforts in keeping the US ahead in AI Technology. This week, we take a look at the race between the US and China from the perspective of AI research output.

A new analysis by the Allen institute for AI shows that China’s AI research publications are rapidly increasing and “the two nations will produce an equal share of top AI publications by 2020”. It is not by coincidence that the Chinese government announced in 2017 a new AI strategy that aims to rival the US by 2020. Similar to DARPA’s playing a key role in the US’s evolving AI strategy, China’s defense ministry is also investing deeply in AI innovation, according to a study by the Center for a New American Security.

Quantity does not always come with quality. The quality of a research publication depends on its citation count and publication venue. On the citation count, the data from the Allen Institute’s aforementioned analysis shows that AI research output from China has improved sharply; a jump from less than 5% in 2000 to almost 30% in 2017 in the share of the top 10% most-cited AI research publications. The US share decreased slowly from 40% to 30% during the same period.

However, given the size of China, it would be interesting to know if the citation data takes into account citations by authors from the same country. Issues have been raised about Chinese research institutions’ reputation for low quality and even fraudulent publications.

Anytime it comes to ethics, whether it is about publishing AI research or developing AI products, it is a serious issue because of the highly potential impact of AI in the future. The AI World Society (AIWS) puts AI ethics as a focus of many of the organization’s works. Last year, we produced a set of Ethical Frameworks for AI Norms and Standards, towards building a social model that will make Artificial Intelligence (AI) safe, trustworthy, transparent, and humanistic.

On the other hand, we cannot wait until the quality of China’s AI research output is verified. “The US government needs to do more to support AI research”, said Oren Etzioni, CEO of the Allen Institute for AI.