Sebastian Thrun, founder of Udacity, on Thursday spoke in Beijing about artificial intelligence and autonomous driving. Headquartered in California, Udacity enables its users to learn the latest and hottest technologies in Silicon Valley via its online courses.
Before founding Udacity in 2012, Thrun worked at Google, which in 2009 started an autonomous driving car project under his leadership. Aside from that, the Stanford professor led some renowned projects there, including Google Glass and Street View.
Thrun believes that so far the autonomous driving industry is at an early stage, as there are no commercialized cars yet.
That is true. BMW is scheduled to ship autonomous cars by 2021. Tesla, probably the most aggressive and bold automaker, set the timeline for autonomous driving in commercial use by the end of this year.
However, Thrun explained that progress in self-driving cars will be considerable. Driverless cars learn much faster than humans. Human drivers all have to learn driving on their own and make the same mistakes. For self-driving vehicles, when one car makes a mistake and learns from it, it’s something that can be copied to all cars.
“If you birth a new car, it’s born with the complete wisdom of all previous self-driving cars. If cars maintain this speed of improvement, I would estimate that every 18 months the driving will get 10 times better,” said Thrun.
He also predicted that in the next one to two years we will see pilot models of autonomous vehicles come out, that is to say, vehicles that are not mass produced on a large scale. He expects that the rollout will be in the order of eight to ten years for self-driving cars used by the majority of people.
Thrun has seen China as an ideal land for research in autonomous driving and AI. It is home to a gigantic number of car riding and complex transportation situations. Didi Chuxing, China’s largest ride-hailing app, provides 20 million trips per day.
Few would dispute that artificial intelligence (AI) is at the core of autonomous driving. The AI veteran thinks that the most recent wave of AI is all concentrated around deep learning and machine learning. Thrun stressed that this ability of a machine to learn by itself is what makes the field so full of potential.
“In the past when you programme a computer, you would write very elaborate computer code that would tell the computer what to do. That is very hard to do. With deep learning, instead of telling computers what to do we teach computers what to do,” said Thrun.
(Top photo from Flickr.com)