Researchers at the Centre for Security and Emerging Technology (CSET), a think tank of Washington-based Georgetown University, assessed in an issue brief this month that China's public investment in AI R&D was in the order of a few billion US dollars in 2018, similar to the US’ planned spending for financial year 2020, and not the tens of billions previously suggested in some quarters.

If the lower estimate proves to be accurate, it would undercut some of the key assumptions underpinning calls by the US to step up investment in AI. In February, US President Donald Trump signed an executive order directing the US government to prioritise AI in its research and development spending, months after he reportedly received a memo from then-Defence Secretary James Mattis that suggested the creation of a national strategy for AI to keep pace with China.

US urged to invest more in AI by panel led by ex-Google chief

A report by the National Security Commission on Artificial Intelligence (NSCAI) also declared last month that China is investing more in AI than the US. Google’s former CEO Eric Schmidt, who led the government-commissioned panel, warned specifically that China was ahead in facial recognition and financial technology.

Amid the battle for AI supremacy between the world’s two largest economies, Washington blacklisted eight prominent Chinese AI firms in October over Beijing’s alleged human rights violations of Uygur Muslims and other predominantly Muslim ethnic minorities.

US Air Force General VeraLinn Jamieson projected in 2018 that China would spend tens of billions of dollars in the field: “We estimate the total spending for artificial intelligence systems in China for 2017 was US$12 billion. We also estimate it will grow to at least US$70 billion by 2020,” she said at an Air Force event.

How China’s AI tech founders plan to survive Trump’s blacklist

The US’ annual federal spending on non-defence-related AI research is set to reach nearly US$1 billion, according to a supplement to President Donald Trump’s fiscal 2020 budget request released in September.

The CSET researchers said their findings were preliminary, as they relied on sparse and uneven open-source data as well as many assumptions. However, they added that the rough estimate and the underlying analysis allowed them to rule out higher estimates with “moderate to high confidence”.

The CSET researchers had “low to moderate confidence” in its estimate of China's public investment in AI R&D because significant data gaps prevented them from producing a more precise estimate of China’s spending, especially with respect to defense R&D.

The estimate took into account China’s total public investment in non-defence-related R&D, using spending figures from China’s Ministry of Finance as well as open-source data from two representative public programmes – the National Natural Science Foundation of China and the National Key R&D Programs.

They also estimated China’s military AI R&D spending using line items in MOF’s spending data that did not have descriptions, assuming that this undisclosed expenditure may be related to defence.