The first day of Google I/O has been full of AI-related announcements, with the technology bringing new features and capabilities to Google News, Google Photos, and more. But arguably one of the most important areas for AI is autonomous driving, and Waymo has revealed that the progress made is finally allowing the company to start deploying truly autonomous vehicles.
Phoenix, Arizona has already been the testing ground for Waymo's self-driving cars, but now the company is confident that its self-driving cars are so safe that a human driver behind the wheel is no longer necessary, and, just like with its first autonomous vehicles, Phoenix will be the first location to get these truly driverless cars as well.
To get to this point, Waymo had to make improvements to both detection and behavior prediction. In one of the given examples, the company talked about how its cars can now predict behavior at an intersection, even if the light is green for the self-driving car. For instance, the vehicle can predict that a particularly fast-moving car will run the red light at the intersection, and so it will slow down to let the car go through even though it's not the typically expected behavior from a driver at a red light.
Waymo also revealed that AI has helped reduce the error rate in pedestrian detection by 100 times and has enabled the car's sensors to see past particularly rough weather conditions, such as snowstorms, where snowflakes might overload the sensors with data. AI has made it possible to eliminate that noise and still identify other vehicles on the street.
The news comes at a time when self-driving cars have been involved in some undesired situations but if Waymo's confidence is justified, it might put some people at ease concerning the safety of autonomous driving.
Via: Android Authority