We don’t talk as much as we might about computer autonomy here, perhaps because it’s obvious that the biggest challenge facing interstellar flight is propulsion. But it’s clear that we need computer systems with fully autonomous characteristics on the kind of decades-long robotic missions that might eventually be flown. We’ll want such probes to have human-like traits of curiosity and judgment, as well as repair and maintenace capabilities for the long journey.
It’s interesting to see, then, that Cambridge, MA-based BBN Technologies, which played such a pivotal role in the development of the ARPANET and later Internet protocols, has just received funding to create a so-called ‘Integrated Learner.’ Working to the tune of $5.5 million from the Defense Advanced Research Projects Agency (with a possible total of $24 million over four years), BBN’s first phase effort will be to create a reasoning system that can apply itself to advanced tasks and master principles of the kind we all refer to as ‘common sense.’
“This program attacks one of the biggest problems in AI,” said Mark Berman, vice president, BBN Technologies. “The Integrated Learner will combine traditional machine learning techniques with an AI reasoning system capable of understanding behavior it observes only once. This ambitious goal is necessary because, for many of the current and future complex military tasks that could benefit from automation, there simply are not many examples in existence. Although there has been some research into this area, this will be the first deployed system with the capacity to apply general knowledge and reasoning to a task.”
Up first is research on military medical logistics planning, involving a simulation of the evacuation of wounded soldiers to various hospitals out of the combat theater. But the larger picture relates to building digital systems that can learn complicated tasks and apply the acquired skills broadly. Computer learning isn’t particularly captivating until you see it in context, but go read Greg Bear’s wonderful Queen of Angels (Warner Books, 1990) and you’ll see a malleable, motivated AI system built into the world’s first probe to Alpha Centauri, and learn what happens to a truly adaptive system once it begins to fathom the true meaning of its mission.
