Future wars will be fought by robots - or at least that is how tomorrow's battlefields have been seen for decades in the popular imagination. From Star Wars to the Terminator movies, a range of smart, self-aware mechanical warriors routinely fight alongside, or instead of, human soldiers, while other, less warlike versions help perform essential tasks and keep the troops resupplied.
For decades, too, militaries thought the same, as significant amounts of the world's defence budgets were ploughed into robotics research, with the US leading the charge. Pentagon spending funded everything from remote-controlled 'gunbots' such as the Modular Advanced Armed Robotic System (MAARS) and the free-swimming Mantabot fish drone, to the Big Dog robotic beast of burden, and even a bona-fide android known as Atlas.
This push created a number of significant developments, including bomb squad robots deployed for use in Iraq and Afghanistan, the in-field use of the man-portable RQ-11 Raven reconnaissance drone and a lengthy field evaluation of a Big Dog variant by the US Marines. Shortly before his retirement in 2014, General Robert Cone, head of the US Army Training and Doctrine Command (TRADOC), spoke openly of a quarter of combat forces being replaced by robots by 2030.
Robots were on the rise - and then two things happened.
Don't be evil
Even as General Cone spoke, Google was coming towards the end of a shopping spree that saw the technology giant snap up eight notable robotics companies in less than as many months, culminating in the purchase of Boston Dynamics, the firm behind Big Dog and its kin. Suddenly, the new and semi-secret Google X research facility was offering a new source of high-risk, high-reward funding for aspiring roboticists next to the Defense Advanced Research Projects Agency (DARPA) which had for so long been the only patron with deep enough pockets to afford the work.
While Google's new robot empire was to remain linked to DARPA projects for a year or so by existing pre-acquisition contracts, the firm's rejection of military funding for further work on Atlas made its position very clear. Whether driven by a lofty desire to live up to its "don't be evil" mantra, or the rather more pragmatic attraction of the profits to be made in the lucrative commercial sector, it looks as if Google will not be involved with the robot soldiers of the future. With many of the best makers of robotic arms, legs, wheels, visual systems, sense organs and brains now firmly ensconced in the Google X fold, things look very different.
What killed Big Dog?
Although much more mundane than Google's corporate takeover of the robot world, in many ways the second event was every bit as much of a game changer: after $32m of investment and five years of testing and development, the US Marine Corps finally killed off Big Dog.
Designated the Legged Squad Support System (LS3), the militarised variant had been expressly designed to help carry the heavy loads needed by expeditionary forces, literally easing the burden for human warfighters. While it easily met the requirements of carrying 180kg in a sure-footed way, and covering 20 miles in under 24 hours without refuelling, its downfall was to be the noise it made doing so. Being followed around by a machine that reportedly sounds about as covert as a two-stroke lawnmower is hardly likely to win favour with troops in a combat setting.
According to a spokesperson for the US Marine Corps Warfighting Laboratory, "they took it as it was; a loud robot that's going to give away their position."
So where does all this leave military robotics?
The third offset
"The military have always been in two minds about robots, both remote-controlled and autonomous," says Stuart Russell, professor of computer science and Smith-Zadeh Professor in engineering at the University of California, Berkeley, and one of the signatories to last summer's open letter calling for a ban on lethal autonomous weapons. "But [US Deputy Secretary of Defense] Robert Work's recent comments to the press and other comments made to me by military policy types suggest the US still sees autonomous weapons as a potential 'third offset'."
The US has used the offset principle twice before to ensure its advantage against superior conventional numbers in the Cold War era, by emphasising nuclear deterrence in the early years and by developing GPS, precision munitions and stealth technology towards its close. The Pentagon now hopes advanced robotics could help form part of a third offset strategy - particularly at sea and in the air. Autonomous systems such as the strike-fighter sized X-47B Salty Dog - Northrop Grumman's contender for US Navy's unmanned carrier-launched airborne surveillance strike (UCLASS) acquisition programme - and the 130ft, 140t Sea Hunter robot boat have already established their credentials, and others are set to follow.
According to Armin Krishnan, assistant professor of political science at East Carolina University, whose research interests include military robotics and military ethics, government secrecy, targeted killings and shadow wars, currently manned systems could be made robotic, too - and not just by the US.
"Existing weapons platforms that have already fly-by-wire or other electronic control systems can be easily roboticised," he says. "F-16s may be converted into unmanned fighter aircraft that fly alongside some manned F-22s. The Russians plan to convert their T-90 tanks to unmanned systems; the new Armata tank is rumoured to have been developed for later full roboticisation."
When it comes to new robot designs, despite Google's position at the forefront of AI development and the company's stance on defence work, Krishnan does not think the general trend of building unmanned systems will be significantly affected. "AI is advancing very rapidly and there will be a strong temptation to leverage for the military," he says. "You should not merely think of humanoid bi-pedal robots like the Terminator - robots and AI come in many shapes and forms."
One day, Krishnan believes, that may also include biological robots - possibly cyborged animals, if not humans - and cites the work being done by DARPA on brain implants that can control mental states and thereby behaviour.
There is an obvious appeal for military technologists to go down the biological route; as Krishnan points out, nature can construct machines much more elegantly and more efficiently. A humanoid robot either has to rely on a tether, or use a battery which will at some point run out of power, and often very quickly; a self-sustaining living organism, by contrast, suffers from neither of these limitations. In addition, he says, nature produces more cheaply.
"I am sure a humanoid robot would cost millions of dollars to produce and I don't see how costs could realistically come down much since expensive components would be always required," he explains. "With biotech [you] could engineer animal-human hybrid super-soldiers at a fraction of the cost. Sounds horrible. Well it is. But it is happening."
It is perhaps ironic that in Karel Capek's 1920 play 'R.U.R.' - which introduced the word 'robot' into the lexicon for the first time - the 'robots' are not the kind of electro-mechanical beings we mean today, but rather a form of actual artificial life. Perhaps then, with the rise of synthetic biology, and the concomitant surge in defence funding now being channelled into that field, robotics will simply come full circle, return to something more akin to Capek's original vision, and ultimately prove Krishnan right.