Experimental AI Powers Robot Army

Darpa's Grand Challenge may have looked tough, but it was a piece of cake compared to the challenge facing robots currently being developed by the U.S. Air Force.

Rather than maneuver driverless through miles of rough desert terrain, these will have to find their way into underground bunkers, map unknown facilities in three dimensions and identify what's in them while avoiding detection -- all without any human control.

http://www.wired.com/news/technology/software/0,71779-0.html?tw=rss.technology

:scared:

http://en.wikipedia.org/wiki/Skynet

In fact, it's not a whole lot more complicated than the AI found in most First Person Shooter games.

However, the really, really, really hard part is providing that AI kernel with enough knowledge of it's surroundings to be able to make these decisions. Computer vision still sucks, last I heard. GPS-navigating AIs have done a much better job of "living in the real world", but only if there's nothing unexpected in front of them (they won't see it). Radar, sonar, etc... none of them are fine-grained enough to do the job yet.

But once you get all that environmental/navigational/positional data into the bot, the rest is not that hard to do.

most of what we consider "intelligence" is really perception. Which does not minimize the magnitude of the problem, it just moves it into a different arena. The fundamentals of navigating and manipulating in the real world of sensory clutter and time-limited decision making (and staying alive and uninjured while you do it) -- that is where the action is.

Which is another way of saying "lord, if only I could get my software to be as capable as the average 3-year-old... I'd be both rich and famous."