Hacker News Reader

Comment by epcoa | Hacker News Reader

The person you're replying to never claimed otherwise. However, while decision support is not directly steering and accelerating/braking the car, I am just going to assert it is still driving the car, at least for how it actually matters in this discussion. And the best estimate is that these interventions are "uncommon" on the order of 10ks miles, but that isn't rare.

A system that requires a "higher level" handler is not full self driving.

View on HN

I think the important part is that the remote person doesn't need to be alert, and make real time decisions within seconds. As I understand it, the remote driver is usually making decisions with the car stationary. I'd imagine that any future FSD car with no steering wheel would probably have a screen for the driver to make those kind of decisions.

There's a simple test I find useful to determine who's driving:

If the vehicle has a collision, who's ultimately responsible? That person (or computer) is the driver.

If a Waymo hits a pole for example, the software has a bug. It wasn't the responsibility of a remote assistant to monitor the environment in real time and prevent the accident, so we call the computer the driver.

If we put a safety driver in the seat and run the same software that hits the same pole, it was the human who didn't meet their responsibility to prevent the accident. Therefore, they're the driver.

Agreed!

Which is why an autonomous car company that is responsible and prioritizes safety would never call their SAE Level 4 vehicle "full self-driving".

And that's why it's so irresponsible and dangerous for Tesla to continue using that marketing hype term for their SAE Level 2 system.

In that case, it sounds like "full self driving" is more of an academic concept that is probably past it's due date. Waymo and Apollo Go are determining what the actual requirements are for an ultra-low labor automated taxi service by running them successfully.

Reply on news.ycombinator.com