If they could roll it out to everyone they would do so in an instant. They cannot though because the system is not ready and needs a lot of testing, which is the point the author is making.
It also doesn't seem to help that enabling FSD on Tesla makes it use the visual camera only, which seems counterproductive. The ultrasonic sensors seem to be critically important because visuals can be obstructed easier.
Can you post a link to a video of FSD nearly hitting a pedestrian? Because searching on YouTube, I couldn't find a single one.
The car still uses ultrasonics when parking but AFAIK it has never used them for driving around. Perhaps you meant radar?
It was publicly known that Musk was personally involved with requiring all sensors to be completely hidden from view, even to the point of inhibiting the car's view of its surroundings. This was back when other cars had the large LIDAR installations on top of the roof, which was probably better, but obviously didn't give the regular-car look Tesla wanted.
All I can say to comments like this is "call a friend and get a ride". There are ~23k cars on US roads with this system now. They're not hard to find. And it's really pretty mind blowing. For sure the most fun I've had behind the wheel in 25 years of licensed driving.
The very fact that you are able to cherry-pick failure cases from a mere fleet of 23k cars means the system is by no means ready.
In fact, you/Tesla/everyone involved should encourage people to "try your best to cherry-pick failure cases". What you are advocating is the opposite of being responsible (let's try to ignore the failure cases and focus on when it works!)
>For sure the most fun I've had behind the wheel in 25 years of licensed driving.
I'm sorry but the amount of "fun" you have behind the wheel is not remotely close to being a success criteria for a system named "full self driving".
To be clear: what I was actually saying was not to ignore videos on youtube, but to be responsible about interpreting them. When data sets get large, you need to stop reasoning on the basis of anecdotes (because there will always be single data points to support any perspective) and start making your decisions on the basis of data.
Basically: is FSD beta as deployed any more dangerous than human drivers? I think it's clear to everyone that the answer is "no", for the simple reason that the roads are filled with bad drivers already. In fact the one actually documented accident involving a beta car of which I'm aware[1] was one where the car got T-boned by a red light runner.
Really, I'm being sincere: stop being angry on the internet about this and get yourself a ride. You'll see things differently, I all but guarantee it.
[1] Sort of. It was on twitter and I've lost the reference and might be misremembering.