Autonomous driving companies like Aurora are plunging their driverless algorithms into hyper-realistic virtual worlds to test AI vehicle safety. With this approach, they can create thousands of dangerous scenarios and throw anything at their vehicles - without posing a risk to anyone.

But what kind of driving scenarios do you train a driverless vehicle against? Autonomous vehicle testing company dRISK have captured a million hours of camera data and hundreds of thousands of accident reports - including collisions with portable toilets and picnic tables – to try and answer that question.

00:00 How do driverless cars prove that they can handle any driving situation?
01:10 Aurora’s “hyper-realistic” truck driving simulator
03:06 dRISK is mapping edge-cases
04:42 Will driverless cars always need to train on real roads?

Check out my extended discussion with dRISK if you want to know more about how the company is mapping out risky, rare, and dangerous driving scenarios:    • How To Train Your Self-Driving Car? U...  

I'm George Downs, a WSJ video journalist fascinated by how technology is changing how we get from A to B. If you're interested in the future of mobility or how modern transportation can impact our lives– from EVs to eVTOLs, and beyond – then don't forget to subscribe.

#DriverlessCars #AI #GeorgeDowns