FSD Beta in San Francisco

[bayareaever]

I live in the East Bay and have had plenty of the problems reported here with fsd beta, to the point where it isn’t all that useful near where I live.

Today I drove around San Francisco for the first time and the performance is noticeably better, to the point where it is actually quite useful. Maybe just a fluke - have others in SF had similar experiences? Elon has implied that the fsd SF Bay Area performance is better than in other places. Maybe that doesn’t include the East Bay?

 

[BM3LRPUP19]

I found it to be garbage, dangerous, ignoring large obvious objects, humans, muni at times, or overly cautious then shifting to aggressive. I've had it on steep hills, mission street, dolores, cesar chavez, 18th. Just when I think there was an improvement from the previous version I'm shutting it off to avoid a crash. I don't understand this preoccupation or fantasy about vision. If I had lidar embedded in my head I'd use it at as another sensor at least to cross check if a nothing at all.

 

[Jigglypuff]

Lidar isn't going to solve the problems, though it might help. The real issues with self-driving tech are algorithmic, not sensory. Lidar may maybe one part of the problem easier, but only one.

We'd be much better off if the government stepped in and mandated smart roads and interconnected cars.

 

[bayareaever]

I found it to be garbage, dangerous, ignoring large obvious objects, humans, muni at times, or overly cautious then shifting to aggressive. I've had it on steep hills, mission street, dolores, cesar chavez, 18th. Just when I think there was an improvement from the previous version I'm shutting it off to avoid a crash. I don't understand this preoccupation or fantasy about vision. If I had lidar embedded in my head I'd use it at as another sensor at least to cross check if a nothing at all.

I was mostly driving in western SF where it tends to be a more tame driving environment. I assume that accounts for most of the difference.

 

[aronth5]

Lidar isn't going to solve the problems, though it might help. The real issues with self-driving tech are algorithmic, not sensory. Lidar may maybe one part of the problem easier, but only one.

We'd be much better off if the government stepped in and mandated smart roads and interconnected cars.

In what decade would you expect there would be enough smart roads and interconnected cars to actually improve safety?

 

[thewishmaster]

I found it to be garbage, dangerous, ignoring large obvious objects, humans, muni at times, or overly cautious then shifting to aggressive. I've had it on steep hills, mission street, dolores, cesar chavez, 18th. Just when I think there was an improvement from the previous version I'm shutting it off to avoid a crash.

My experience as well. It's generally fine when there are no vehicles or pedestrians around and you have time to let it figure out some of these things on its own, but otherwise I take over constantly to avoid causing unnecessary problems.

 

[Jigglypuff]

In what decade would you expect there would be enough smart roads and interconnected cars to actually improve safety?

Interconnected cars could happen in 10 years if it were mandated soon. It's a long way off since there are few efforts to make them happen. Smart roads could happen relatively easily. At the next repaving, put a wire in the center of each lane.

Instead, billions upon billions are being poured into mediocre to failing self-driving efforts. We still have garbage speech to text. I don't know why we expect the far harder problem of self-driving is going to be solved anytime soon.