The US National Highway Traffic Safety Administration has decided to officially recognize Google’s autonomous driving system as a valid driver in the US. This marks the first move of the government moving towards changing the definition of a driver away from being a human.
For Google Self-driven cars, the machine/software is classified as the driver, and not the human/owner of the device. This potentially shift the liability away from the owner of the car, and onto Google, or the company that develops the autonomous driving software. For this reason, Google will need to make their artificial intelligence near-perfect before their cars or software can reach the market, as the potentially liability and cost from errors resulting in injury or death would be enormous.
Enabling AI Software and self-driving devices to be qualified as drivers marks an important milestone for the development of self-driving cars. By fitting within the definition of existing laws, it can increase the freedom of owners of the self driving cars. For example, laws regarding texting while driving or drinking while driving would ideally not apply to those that are in a fully automated car, as they have no control over the vehicle, and are not considered the driver.
An estimated 94% of car crashes are a result of human error, including poor decision making, bad reaction time, and impairments such as lack of sleep or alcohol. If the artificial intelligence engine used by Google reaches near-perfection, it can reduce the number of accidents substantially. In addition, 2% of crashes are attributed to vehicle error. A very significant portion of these “vehicle errors” are actually caused by human errors, such as not getting the proper maintanence on tires, brakes, etc. Ideally, a smart self-driving car could detect potential deterioration of parts through the way the car’s motion reacts to input, and notify the user to arrange a repair, eventually locking the user out for their safety if the conditions of the vehicle become too dangerous to drive.
The 2% of crashes resulting from environmental conditions can also be partially attributed to human error. Slick rodes, glare, and other abnormal conditions can be recognized by the driving software, and the software can drive more cautiously according to the conditions.
Google is not the only company to pursue Autonomous driving. Countless automobile manufacturers have pumped R&D funds into this technology, and NVIDIA recently announced their NVIDIA Drive PX2 device that utilizes deep learning to identify patterns of correct driving and visual recognition, and utilize it to make the correct decisions and properly map out the driving scene in a virtual 3d space.