You're all too skeptical of super-duper self-driving cars, apparently

Easier Said Than Done

"It's not clear how regulation and enforcement will happen," said Vasily Tarasov, a software engineer at Atlas Informatics who recently adapted Comma.ai's open source self-driving car technology to work with his 2017 Chevy Volt, in a phone interview with The Register.

He pointed to online videos of people testing the limits of Tesla's Autopilot and Uber's attempt to claims its Level 3 cars in San Francisco, California, did not need a permit as examples of the kinds of challenges regulators will face.

Tarasov said he expects people will come to accept self-driving technology gradually, as they encounter it over time. But in the near term, he sees a need for more data about how Level 2 systems perform.

"You have to have widespread deployment of Level 2 to gather enough evidence that your system is actually safer than human driving," he said. "Only after you have the statistical evidence can you start negotiating with insurance companies and regulators to formalize this kind of situation."

Tarasov said he hadn't had any close calls while testing Comma.ai's open source system, which "has very good safety rules built-in." He said he hoped more people will contribute to the project.

"I had hoped the community of Comma.ai users would evolve to support more cars," he said, noting that so far developers have added support for only two extra cars. "It's not happening as fast as I expected, but it's a step forward." ®