Hacker News Reader

Comment by Animats | Hacker News Reader

Maybe not. The "second-gen" autopilot hardware is just more cameras and a new radar. No LIDAR yet. Tesla is struggling to get back the performance of their own first-generation autopilot, hopefully without the "crashes into stationary objects partially blocking lane" feature.

Tesla is at least one or two more hardware upgrades from Google/Waymo level performance.

View on HN

I've yet to see one Google/Waymo car doing 90mph on the highway with Lane Keeping Assist Level 3 Autopilot. Not ten, not five, or even one. Tesla has thousands of them doing it as you read this.

and how does that make their computer vision hardware any better?

Cool. Where can I buy a Google / Waymo car? I will be happy to.

Sometimes shipping is a key.

The point on shipping is appropriate... But just for clarity: of all the autonomous cars companies that ultimately "succeed", you will probably not be able to purchase cars from at least half of them. You'll only be able to purchase rides.

How are you so sure of this? Maybe yes, maybe not. I think there is a merit to owning you own car and the traditional model might stand up quite well.

I know a lot of people at a lot of robotics companies (including autonomous cars companies), and have spoken at length with them about the ultimate business models.

Besides, it makes sense when you consider that most personal vehicles have sub-5% utilization. It's more capital and environmentally responsible to have shared assets rather than underutilizing assets with high embodied costs.

What may make sense to you and what people end up doing are often wildly different.

It's easier to make a car able to drive in a single city. Also, companies like Lyft could deploy manually driven cars for "hard" bookings.

Developing an autonomous car for Lyft is easier than developing it for a normal person.

Reply on news.ycombinator.com