Robot Cars Can Teach Themselves How to Drive in Virtual Worlds

credit NPR

Ever since watching The Matrix, I've dreamed about downloading knowledge into my brain, without actually having to study things in real life.

Alas, that's not happening anytime soon, at least not for us humans.

Self-driving cars, on the other hand, are another matter.

Normally, self-driving cars learn to drive by actually going onto real-world roads. Obviously this takes a hell lotta time and a hell lotta babysitting by bored out-of-their-minds human engineers, who're there in case an accident happens.

Recently, scientists are starting to realize that real-world driving may not be necessary to train self-driving cars. Instead, virtual worlds assembled from Google Street Views or even Grand Theft Auto may be used to train the car's software. This means that future autonomous vehicles may be able to learn how to drive by performing simulations in their "heads",  without ever leaving the garage.

In this week's article for Singularity Hub, I look into how these virtual worlds are propelling self-driving car technology forwards by leaps and bounds. Now if only I can speed up my guitar learning by performing simulations in my head...