The stakes are high. An average of 35,000 Americans killed every year. Hundreds of thousand injured, many, with lifetime injuries, chronic pain and permanent disabilities. $450 billion every year to be saved.
All that needs to be done is to perfect self-driving cars.
While the founder of Google projects that his companies self-driving car will be on the market by 2017, it is expected that that will only be in limit usage and special applications. It may be more than a few years before a Google autonomous vehicle is parked in your garage, waiting to whisk you to your work in Philadelphia, down to the beach or up to the Poconos for a weekend.
Driving is more complex than it may appear. Of course, the more closely you examine everything that goes into driving, the more frightening it really becomes. And Google’s researchers are finding that programming a vehicle to be safe is very complex.
With the current models, in an effort to remain safe and avoid car accidents, the cars sometimes become so perplexed as to be paralyzed. Researchers sometimes have to take the wheel and drive them back, as the computer becomes overwhelmed by “bizarre” surroundings, such as a woman chasing a duck with a broom, which so confused the computer that it shut down.
Other problems include becoming a “doormat” for every other driver. They need to program them to function in the “social environment” of driving, which means they need to develop “body language” that communicates to other drivers.
One reason why 90 percent of car accidents are the result of human error is that it is such a complex activity, and because most of us learn by doing, extracting the “grammar” of driving and translating into code for a car could remain a work in progress for many years.
Toptechnews.com, “Google’s Self-Driving Cars Encounter the Bizarre,” Steve Johnson, November 17, 2014