>> it has better eyes than you do.
Actually it doesn’t. According to Google they don’t have the resolution to detect something as small as a squirrel. The sensors also don’t function in fog (I suppose we don’t either) and I suspect they would have problems in heavy rain rain and snow as well. In my system, we used to have huge problems in extremely heavy rain. You can see through a snow storm, but Google cars will think they are stuck in a box. Other issues are that the cars can’t understand gestures of a police officer directing traffic. I also suspect that these vehicles will take you right into a flooded underpass at full speed without detecting a problem.
Eventually we’ll get around the sensor issues. Things like femto-photography and other neat stuff will get us to a working system, but not in the time frame Google’s Urmson is quoting.
Because of liability issues, the computer has to be perfect. Ask Google about their timeline after the government is through with GM. If you get beyond Google’s smoke and mirrors, Googles cars aren’t even close to being as good as a mediocre driver. While Google states that the vehicles have great safety records, talk with them a little further and you find out about the numerous times they’ve had to take over for the vehicles.
The “ball rolling from between cars” is only one scenario to illustrate driver intuition. There are loads of other scenarios. These scenarios illustrate a fundamental flaw in the AI in current autonomous driving implementations.
Here is a excerpt from wikipedia that describes what I’m getting at:
“Human beings solve most of their problems using fast, intuitive judgements rather than the conscious, step-by-step deduction that early AI research was able to model. AI has made some progress at imitating this kind of “sub-symbolic” problem solving: embodied agent approaches emphasize the importance of sensorimotor skills to higher reasoning; neural net research attempts to simulate the structures inside the brain that give rise to this skill; statistical approaches to AI mimic the probabilistic nature of the human ability to guess.”
There is work being done in this area:
That being said, I have few issues with systems that assist and enhance a drivers ability. Lane keeping systems, active cruise control with start/stop capability are here now and will work fine, but you still need a human driver keeping an eye on the store.
If I seem a little extreme and overly concerned, there’s a reason. A couple of weeks ago my daughter took a plane trip. From the time she rolled away from the gate to the time she arrived, she was being protected by software I helped design and code. I hadn’t anticipated that moment when I wrote it, but I was really comforted knowing the effort we put into the products. I’d expect the same level of quality from an autonomous vehicle system. Code it like your kids life depended on it. Your kid could be the one chasing the ball. Talk to WWII bomber pilots about Boeing engineers over designing the B-17.
What’s next, check this out:
Source : http://www.thetruthaboutcars.com/2014/05/qotd-a-robot-car-that-kills-you/580