Hacker News Reader

Comment by browserman | Hacker News Reader

Tesla's AI engineers can't even solve their own problems, why do you think they could solve Twitter's?

They have solved much harder problems than spam bots.

You mean self driving, which they promised would be ready over 5 years ago and still barely at L3 level? Meanwhile Google engineers have actual L4 level self-driving in both Phoenix and San Francisco (by your logic, they are better than Tesla engineers), yet have not managed to solve Youtube's spam problem.

>Meanwhile Google engineers have actual L4 level self-driving

They don't have L4 self-driving they just made a contrived railway system for their cars. The second the car is out of the hardcoded route it's not a self driving car anymore.

>The second the car is out of the hardcoded route it's not a self driving car anymore.

So what you're saying is that at least the car is self-driving sometimes, whereas the Teslas never are. That's your pro-Tesla argument?

You can get in and drive around in a Waymo taxi with no driver and yet Tesla fans are still claiming it's not self-driving, while Tesla is. Hilarious.

First off, it's a hard coded area, not "route". They're not buses. A limited area is part of the definition of Level 4. What you're thinking of is Level 5, which is being able to handle any area and situation.

There are also reasons beyond capability for the cars being limited to an area. One is legislative, they literally are not allowed to offer service outside a given area. Another is maintenance, the cars need to be within range for their team in case of emergencies or accident. The cars have been tested in many cities outside those two cities, but offering a user-facing service has a lot more barriers.

So it dosen't have a single route, it has a couple of them. Mapping the whole area beforehand then driving in it is not "self" driving. You don't go out and carefully map each square inch of the road you're going to take before driving your car on unknown roads do you know?

"Spam bots" on their own aren't a hard problem. The hard problem is a very restrictive set of constraints on the solution space, e.g. "can't inconvenience or increase barriers to posting for legitimate users in any way."

This is the last place I would expect to see naive hero worship of Tesla's autopilot software, of all things.

I think they all came out of the woodwork after the last article.

Solved is a strong word here. They've certainly made a lot of progress

Did they stop killing people by crashing into stationary barriers yet?

Reply on news.ycombinator.com