Trump administration sounds alert as China forges ahead with AI

Washington

PRESIDENT Donald Trump's recently announced artificial intelligence initiative does not include the word "China." But that country's progress in the race to machine-learning supremacy has prompted calls for the United States to start running faster. Though Mr Trump's plan is light on actual planning, the Pentagon released its own report last week that provides more detail. Taken together, the documents are promising.

Mr Trump's order directs agencies to assess their spending, reprioritise existing funds towards artificial intelligence and consider that priority in their upcoming budget proposals. It also opens government data to researchers and the private sector.

And it kick-starts a process for departments to consider how they might regulate machine-learning applications in their purview. Exactly how these things happen is up to agencies - but at least the President is asking them to start thinking, which so far few seem to have done. The Defense Department is an exception.

The need for increased AI investment in defence is urgent. Researchers see machine learning as the electricity of the 21st century, with the potential to transform life across sectors. But US innovators are much more likely to set their sights on, say, crafting autonomous vehicles than they are on aiding the US military. Google decided not to renew its contract with the Pentagon for AI used to analyse drone videos last year, when employees objected to involving themselves in "the business of war".

In China, things are different. If the military wants a company to build a model for it, officials need merely extend the order and provide the data sets, which are also easier to come by in a surveillance state than in a free society.

The exact amount China will spend on artificial intelligence in the coming years has not been made public, but estimates run in the tens of billions of dollars. Individual cities are committing US$15 billion or more on their own.

Challenging China will certainly require more money than the meagre US$1.1 billion the US commits yearly in unclassified spending on machine learning. Those opened-up government databases are also an important step. But to really compete in AI, the government must also woo top talent, such as those sceptical Google engineers. This is where the Pentagon's report is most compelling: The department emphasises the need for ethical and transparent frameworks to govern artificial intelligence. Whatever guidelines officials develop will be crucial in persuading the industry to work with the government.

The goal of matching China on machine learning is not to be more like the Chinese. It is to ensure that the rest of the world is less like them, by setting the terms for the proper use of a powerful technology in areas from defence to law enforcement to healthcare. WP