quote: For years we heard lots of automakers pay lip service to the concept of autonomous driving, with some even spending a good deal of money and time investigating it. GM was perhaps foremost in the field prior to its bankruptcy. Over a half-decade ago, back in Jan. 2008 we road along in a modified 2008 GM Tahoe which won the DARPA's (Defense Advanced Research Projects Agency) 2007 Urban Challenge. As the car zipped around the obstacle course in sunny Las Vegas, Nev. it quickly became apparent why the vehicle won the smart car challenge -- it was pretty good at avoid collisions... really good, in fact. As a demo, a second test driver would cut off the GM car or otherwise block its path. In every case the GM vehicle knew what to do, performing better than many human drivers would in such a case.
quote: The system is designed to ease the driver’s workload on freeways only, in bumper-to-bumper traffic and on long road trips; however, the driver’s attention is still required... because the system will have operational limitations based on external factors such as traffic, weather and visibility of lane markings. When reliable data is not available, such as when there are no lane markings, the system will prompt the driver to resume steering.
quote: But don't look at GM's solution. Look at Google's. It can't drive faster than 25mph. There aren't that many accidents at those speeds, and collision avoidance is pretty simple. Let's bump this car up to 65mph and put it on an Atlanta highway. Or send it to the Bronx and have it navigate the Grand Concourse or Cross-Bronx Expressway during rush hour gridlock. Pretty sure we'll see accidents.
quote: I don't know if humans are better drivers than software, but I DO know that our current AI technology sucks, and humans are far better at decision making.