Keep in mind that the GC was designed to pull new ideas out of the public. DARPA did not allow anyone already taking public funding (big research labs, many big companies) to participate. So most of the groups that have worked on this problem for years (with DARPA funding) suddenly couldn't participate. The result was a set of systems that were built without (much of) the benefit of state-of-the-art research.
I think the idea of the GC was good, but this was the wrong task. Autonomous vehicles are a hard problem, but the focus isn't really on the AI anymore, it's on the engineering. Path planning and obstacle avoidance are two old AI warhorses. They still need work -- but they've been carefully examined for years. If you're going to look for outside-the-box solutions to hard problems, you should use problems that are new to the field (robust, general manipulation, perhaps).
I don't want to put down the efforts of any of the GC teams -- I think it's great that they're working on it, and I wish them luck in the future. But this was a contest where all the most qualified teams couldn't participate. So is it fair to say that this is the best that AI can do? Not even close.
the page in question describes two final projects for an introductory mit graduate class on robotics (i was the ta). while the students in question did very well, their robots aren't "research" any more than a napkin sketch of some fruit is "art". don't worry -- you'll still be able to get in your door when you get home.
and, just for the record, they're not at the media lab. damn negroponte and his imperialist media empire!
Slashdot Mirror
I think the idea of the GC was good, but this was the wrong task. Autonomous vehicles are a hard problem, but the focus isn't really on the AI anymore, it's on the engineering. Path planning and obstacle avoidance are two old AI warhorses. They still need work -- but they've been carefully examined for years. If you're going to look for outside-the-box solutions to hard problems, you should use problems that are new to the field (robust, general manipulation, perhaps).
I don't want to put down the efforts of any of the GC teams -- I think it's great that they're working on it, and I wish them luck in the future. But this was a contest where all the most qualified teams couldn't participate. So is it fair to say that this is the best that AI can do? Not even close.
the page in question describes two final projects for an introductory mit graduate class on robotics (i was the ta). while the students in question did very well, their robots aren't "research" any more than a napkin sketch of some fruit is "art". don't worry -- you'll still be able to get in your door when you get home.
and, just for the record, they're not at the media lab. damn negroponte and his imperialist media empire!