Daniel Wagner

Opinion | As US, China and Russia fight for AI supremacy in new hyperwar battlefield, will good sense prevail?

In this ‘who dares wins’ race, the US has a lead in deep-learning algorithms but China and Russia are making greater strides in applications, unencumbered by privacy or rights concerns

Reading Time:3 minutes

Why you can trust SCMP

Daniel Wagner

Published: 11:30am, 22 Jul 2019Updated: 11:06pm, 10 Dec 2019

Will artificial intelligence (AI) ever possess the nuance, intuition, and gut instinct necessary to be a good military or intelligence analyst? The jury will be out on that question for some time, but the world’s leading governments continue to develop AI with military applications under the presumption that, eventually, machines will possess that ability. China, Russia, and the United States are in a race to determine which country will dominate the AI landscape first.

China and Russia have had a free hand in applying the technology, unburdened by concerns about privacy, civil rights or degrees of presumed acceptability. China has excelled at developing AI via a vast number of well-trained computer engineers who programme machines using an unrivalled amount of data, produced by the country’s 1.4 billion people. And Russia has little hesitation about deploying AI in any number of instruments of war, including hypersonic weapons.

In contrast, the US sees AI principally as a national security tool, to be employed on the battlefield or to thwart terrorist attacks, and trails both countries in some aspects of the development of deployable machines. The manner in which it allocates and spends defence dollars tends to be slow and cumbersome – not conducive to speed or efficiency. That has hobbled the US in reacting swiftly, effectively and proactively in response to some cyberattacks, and has thwarted the development of cutting-edge AI tools.

That said, US researchers have trained deep-learning algorithms to identify Chinese surface-to-air missile sites hundreds of times faster than their human counterparts. The algorithms helped individuals with no experience of imagery analysis to find the missile sites scattered across nearly 90,000 square kilometres of southeastern China, for example. The neural network used matched the 90 per cent accuracy of human experts while reducing the man-hours needed to analyse potential missile sites from 60 hours to just 42 minutes. This has proven extremely useful, as satellite imagery analysts were drowning in a deluge of big data.

AI and robotics – the forces ushering in the era of “hyperwar” – already allow for asymmetric responses that are inexpensive, resilient and globally scalable. AI technologies such as natural language-based dialogue systems consume enormous amounts of information to augment human operators in non-combat situations, such as for maintenance and the remediation of equipment. Such capabilities will eventually be augmented by reality-based information-delivery technologies in combat scenarios.

At the operational level, commanders will be able to “sense”, “see” and engage enemy formations far more quickly by applying machine learning algorithms to collect and analyse huge quantities of information, and direct swarms of complex, autonomous systems to simultaneously attack the enemy.