Using FSD for 3 months - Is my experience typical?

[PianoAl]

But the software can't learn from our disengagements except through new releases, right?

One reason I asked that is related to a particular lane drift problem. At one location, the car frequently, but not always, went into a right-turn-only lane inappropriately.

Then, it stopped making that error. Ten times in a row, it behaved perfectly. Problem solved? Did it learn from my disengagements?

Well, yesterday it made the error again.

So, a cautionary tale for those who write, "I installed the new update and saw a day and night difference in behavior." This robot's inconsistency can lead to unwarranted conclusions.

 

[mpw2300]

I've been trying FSD in my 2022 S for the last 3 weeks. I'm not renewing it beyond this first month. At best, it's half-baked software. At worst is a risk to myself and anyone else on the road. It's propensity to move into the passing lane when there is no traffic ahead is annoying and dangerous when faster traffic is approaching in the passing lane. It constantly gets confused with wide off ramps and does not handle onramp merging particularly well. I'll be glad to go back to Auto Pilot, as this feature is more secure and useful.

 

[skshimer]

It has been interesting reading this string of posts. I have not used FSD and would not find it worth the money. I do frequently used the AP on limited access highways and really think it is a stress reliever as mentioned in this thread. That being said, it appears from ads that both Ford and GM’s AP have passed Tesla’s.

 

[cbdream99]

My experience with V12 FSD is quite positive, before that it always go into the turn lanes in the city wrongly making it useless and dangerous. I have been using it for door to door transport without any interventions that involve both city and highway, and handle it quite close to human. Of course, it is not 100% flawless and sometimes it fails and sometimes I just feel uncomfortable how it handles the situation and take over (if I let it goes, it most likely works fine but I am not that brave yet). YMMV depends on where you are, I am in the Dallas metroplex area and with the always constructions and crazy drivers, it handles most situations correctly, likely more training data from my area.

 

[robl45]

Honestly I think its cool tech, but is it worth 100 a month? I don't know. I'm thinking maybe 50 a month in the state its in. Its had a number of mistakes. Some excusable and some real bad. Honestly I think it should be free, because we are beta testing it. Problems so far on a supposed complete system.

1. Misses the up/down poles going into my development--maybe excusable, but it should be able to drive to the house.
2. literally was about to tbone a uhaul the other day.
3. multiple times it has taken way too long to get over to get on the highway/get off the highway and basically when it doesn't make it, it just gives up.
4. tried to go down a one way street.
5. freaked out getting off the highway and tried to turn back on for some weird reason.
6. A number of times it goes too quick coming off/on highways and almost skids off the road. Its better than navigate on autopilot because that 100% would have gone off the road.

This is what I'm supposed to pay 100 dollar a month for? It makes my drive of 100 miles round trip each day more fun, but I shouldn't be paying to fix this product.

 

[Goose66]

Just on a lark, I tried the new autopark yesterday. I was in no particular hurry, our deck has newly painted and well marked spaces, and there were no other cars around. I pulled forward and selected a space just behind me that it could back into in one simple turning motion.
It couldn’t have been worse. I mean if they had literally tried to write code to make people think the car was completely incapable, they couldn’t have made it perform worse. It took 90 seconds to back in, with multiple jerky start and stops and pull forwards and frantic wheel turning. I was afraid if anybody saw me I was going to have to jump out of the car and make sure they knew I wasn’t showing up to work drunk. How completely embarrassing must it be for a company that has touted themselves as the leader in autonomous driving for going on 9 years now to only be able to handle such a simple and basic function this poorly?

 

[Dewg]

Just on a lark, I tried the new autopark yesterday. I was in no particular hurry, our deck has newly painted and well marked spaces, and there were no other cars around. I pulled forward and selected a space just behind me that it could back into in one simple turning motion.
It couldn’t have been worse. I mean if they had literally tried to write code to make people think the car was completely incapable, they couldn’t have made it perform worse. It took 90 seconds to back in, with multiple jerky start and stops and pull forwards and frantic wheel turning. I was afraid if anybody saw me I was going to have to jump out of the car and make sure they knew I wasn’t showing up to work drunk. How completely embarrassing must it be for a company that has touted themselves as the leader in autonomous driving for going on 9 years now to only be able to handle such a simple and basic function this poorly?

And yet I get them a round of applause, as the new autopark works much smoother for me, usually doing it in one move, sometimes two, vs the older one that would take 2-4 moves to complete. Sorry your car doesn't work right. :-(

Have you tried picking a spot ahead of you instead of behind you? See if that helps.