There are more than 2000 ground robots fighting alongside flesh-and-blood forces in Afghanistan, according to Lt Col Dave Thompson, the Marine Corps' top robot-handler. If his figures are right, it means one in 50 US troops in Afghanistan isn't even a human being. And America's swelling ranks of groundbot warriors are being used in new, unexpected, life-saving ways.
But there's one small problem: however numerous, these rolling and crawling robots are still pretty stupid. And there's not much hope they'll get any smarter anytime soon.
Groundbots first made inroads among bomb-disposal units. The human bomb-techs could take cover and steer in a remote-controlled Talon or PackBot to disable a dangerous explosive device. But a third of the 1400 fresh ground bots deployed to Afghanistan in 2009 and 2010 weren't for EOD, Thompson pointed out during a presentation at a Washington, D.C. trade show "Robots are not just for explosive ordnance disposal teams anymore … They [ground troops]are using them in ways we never expected."
For instance, at least one unit sent its four-wheeler-size M-160 - a tracked vehicle fitted with a "flail" for detonating buried mines - to scout ahead of a (manned) Husky bomb-detecting vehicle. Route-clearing for the route-clearer, if you will. Thompson played a video that "showed a powerful roadside bomb destroying the M160," National Defense reported. "That would have otherwise been the Husky and its occupants," the magazine helpfully pointed out.
Bots are also being used to inspect vehicles approaching checkpoints, Thompson explained. Many other uses for unmanned ground vehicles are classified, he added.
Thompson seems bullish about groundbots' prospects, but other military robotics engineers have expressed their disappointment. For all the growing popularity and utility of America's robot soldiers, they're still way too dumb to do much of anything on their own. Demonstrations of groundbot technology "have abounded," Dr Scott Fish, the Army's chief scientist, said at the same D.C. event. But military researchers still "don't know when it is we can deliver… serious autonomy."
And about 2000-robot figure. We don't doubt that 2000 robots have been delivered to the war zone, but how many of those are sitting on a shelf in the battalion supply room because they're too flimsy for combat, too dumb to contribute to the fight or simply unneeded?
In other words, for the foreseeable future groundbots will be limited to missions where human operators can closely supervise them - unlike airbots, some of which can already fly many missions with very little human guidance. We won't be seeing totally robotic ground convoys or robot snipers any time soon.
That's because on the ground, "even a twig in the road is an obstacle," one Army researcher explained for my 2008 book War Bots. Whereas, in the air, robots can usually move fairly freely without colliding with anything. It's fair to say this "sense and avoid" problem is now the major focus of military groundbot developers. Despite some promising results at a major Marine Corps test last year, the Pentagon still doesn't know how to train its terrestrial bots to be truly independent.
But that just limits the breadth of ground bots' potential, not the depth. For the missions they're good at - bomb-disposal, checkpoint duty, scouting in close cooperation with human troops - ground bots are really, really good. And there's still plenty of room for more supervised bots to handle more of these particular duties.
Just don't expect them to do your thinking for you.
Wired.com has been expanding the hive mind with technology, science and geek culture news since 1995.