Few days ago Bloomberg's Ashlee Vance wrote about George Hotz who hacked his car and taught how to drive by itself.
"In the month before our first drive on I280, Hotz spent most of his time outfitting the sedan with the sensors, computing equipment, and electronics. Once all the systems were up and running, he drove the vehicle for two and a
half hours and simply let the computer observe him. Back in his garage, he downloaded the data from the drive and set algorithms to work analyzing how he handled various situations. The car learned that Hotz tends to stay in the middle of a lane and maintain a safe distance from the car in front of him. Once the analysis was complete, the software could predict the safest path for the vehicle. By the time he and I hit the road, the car behaved much like a teenager who’d spent only a couple of hours behind the wheel."
 | |||||
| Inside the car. Credit Peter Bohler for Bloomberg Businessweek |
At the beginning the car was unsecure and unstable. Then it learnt how to follow road lines, and be more careful when a cyclist was riding.
What amaze me is the capability of a computer to learn without putting specific rules of behavior inside. Hotz's software is made of 2,000 lines of codes. (A book-sharing app I had in mind requires around 500,000 lines). He says that if/then statements - the rules the computer will follow when he encounters that certain situation - are imprecise and unreliable. In this way the car can learn a personalized way of driving, based on the single driver teaching and experience.
Go forward George!
(Now, my question is: How can we use the Deep Learning technology to automate a mechanical, repetitive task? Proposals welcome)
