Wiki FSD’s Earliest Adopters Still Waiting

[fcastrillo86]

Although AP prevents from dinging but I wouldn't trust it to risk my life.. I used to activate AP when onramp, but this proves to be dangerous and can be deadly. Once I activated AP, it speeds up to 70 rapidly but without the ability to see ahead, what's coming behind from the highway traffic, many times I have to brake or usually AP disengage, once I almost hit by a huge 18 wheelers traveling at 80mph. Having to avoid getting dinged F2C, merge into fast traffic, watching front, back, left, and watch for AP behavior at the same time is just too much. Now, I just try to go out at slower traffic time, find another longer onramp, stay below 50, merge and then activate AP.

On the highway, I constantly have to adjust AP speed to avoid having to use the brake ... far too many times when traffic slows down to 20mph, jerks will suddenly merge into the slightly faster lane, and it's not comfortable to wait/trust AP for serious hard brake... this is just so stressful and certainly not enjoyable. On the city street, I just don't trust AP reacts fast enough ... street is not always straight and flat, I got quite a few HB because to avoid hitting the "stop half of the head out on the street", "dash to make a left turn or stop halfway in the middle", no to mention the aggressive turns for just tiny faster left turn on flashing yellow signal.

Although I maintained a 100 SS and Tesla insurance saved me lots of money, sometimes I doubt whether this is all worth the stress.

I should clarify from my last post that you absolutely need to be aware of your surroundings when engaging and using AP. If AP is accelerating too much to where it might brake check itself for approaching another car too fast, lane merges with nearby cars, etc., then I will always take over. Usually it's enough for me to temporarily steer and re-engage AP. But sometimes, like in the case of sudden slow/stopped traffic, I'll disengage and manually brake early so that I don't have to deal with feeling the hard braking.

That said, it's helped me immensely to just know that AP in its current (and probably future) iterations is a driver assistance tool. Knowing its limitations and where it'll do well vs. freak out is key in using AP comfortably for me. If NoA wants to switch lanes, but I see another car rapidly approaching, I'll wait til the car passes to clear the wheel nag/confirmation so that I don't have to deal with potential swerving behavior. Or in the rare case that it feels like the car isn't going to stop at the line, I'll finish braking myself, knowing that I will most likely avoid a hard braking safety score event because of the 3-second dead time after AP is disengaged.

 

[swt2c]

You certainly are. But currently nothing in the wide release version of FSD is different with or without the camera upgrade.

I wonder about that. My gut tells me that the production FSD features *probably* perform better with RCCB cameras, if for no other reason than Tesla probably doesn't spend a lot of resources testing the RCCC / HW3 configuration, given that it's such a tiny percentage of the global fleet.

 

[Knightshade]

I wonder about that. My gut tells me that the production FSD features *probably* perform better with RCCB cameras, if for no other reason than Tesla probably doesn't spend a lot of resources testing the RCCC / HW3 configuration, given that it's such a tiny percentage of the global fleet.

The only difference is the color filters.

What performance difference reason do you think there'd be for lane change not being accurate because of what color another car is?

 

[swt2c]

The only difference is the color filters.

What performance difference reason do you think there'd be for lane change not being accurate because of what color another car is?

Traffic light color detection is the obvious example that comes to mind. But also lane line detection perhaps. Emergency light detection?

 

[Knightshade]

Traffic light color detection is the obvious example that comes to mind. But also lane line detection perhaps. Emergency light detection?

Is there any evidence of any performance difference at all for any of those?

 

[fcastrillo86]

The only difference is the color filters.

What performance difference reason do you think there'd be for lane change not being accurate because of what color another car is?

Not sure how accurate this is, or if my memory serves correctly, but i swear I saw a post/tweet from @verygreen or @greentheonly (twitter) where he mentioned that Tesla is only collecting training data now from RCCB cars, and that RCCC cars were no longer uploading anything for training of the new neural nets. I'm at work so no Twitter access. But if I find it later on I'll circle back and post here

 

[valem]

Is there any evidence of any performance difference at all for any of those?

It definitely had a hard time differentiate between flashing yellow and flashing red, mostly at night.

 

[Knightshade]

It definitely had a hard time differentiate between flashing yellow and flashing red, mostly at night.

Interesting- here Green seems to suggest the older cams let more light in so might work better (overall) at night, though maybe not for that specific situation- which might mean it's both better AND worse than the newer ones depending on situation.

https://twitter.com/x/status/1451593176716910611

 

[tonglaji]

It definitely had a hard time differentiate between flashing yellow and flashing red, mostly at night.

^^^ this and when the lights turn yellow. I've had multiple times that the car slammed on the break just before it passed the line when the lights turned yellow.

 

[swt2c]

Is there any evidence of any performance difference at all for any of those?

I don't have any, but it would be difficult for anyone but Tesla to come up with any hard statistics (outside of perhaps finding two identical vehicles, one RCCC and one RCCB and conducting a controlled experiment).

 

[valem]

Elon last night at Cyberrodeo said that they plan to go wide release to all that paid for FSD, yet they have not added anyone in the US in over 3 months.
My first thought was, why do they keep torturing everyone with the safety score to get early access?
All this is very upsetting; if true then I think they are taking advantage of all these people using the safety score so they can collect data with the sole purpose to fine tune it for the Tesla insurance.

 

[Knightshade]

Elon last night at Cyberrodeo said that they plan to go wide release to all that paid for FSD

For the beta? (because obviously the finished release will go wide when it's ready)

If so, do you have an exact quote? I don't recall him quite saying that- and it seems a lot of the "pain" in this thread has been people not quite getting Elons quotes exactly right.

My first thought was, why do they keep torturing everyone with the safety score to get early access?

This is why I mean by folks being dramatic here.

Go see what's going on in eastern Europe then come back here and tell us about your torture.

All this is very upsetting; if true then I think they are taking advantage of all these people using the safety score so they can collect data with the sole purpose to fine tune it for the Tesla insurance.

if only you could opt out or something!

 

[fcastrillo86]

For the beta? (because obviously the finished release will go wide when it's ready)

If so, do you have an exact quote? I don't recall him quite saying that- and it seems a lot of the "pain" in this thread has been people not quite getting Elons quotes exactly right.

I can't open YT at work, but check the part of the presentation where he briefly talks about FSD. He asks the crowd if anyone has used the FSD beta, hypes up the 10x safer than human driver, etc., then says something to the effect of "we will go wide with FSD Beta for cars with FSD in North America this year."

 

[Cheburashka]

"we will go wide with FSD Beta for cars with FSD in North America this year."

This isn't surprising. He's been saying this in various variations since 2017... I'd be surprised if he didn't say this.

 

[Cheburashka]

For the beta? (because obviously the finished release will go wide when it's ready)

If so, do you have an exact quote? I don't recall him quite saying that- and it seems a lot of the "pain" in this thread has been people not quite getting Elons quotes exactly right.

There's a bunch of news articles of him saying exactly that: Tesla release FSD beta to North American customers in 2022 – TechCrunch

“The Tesla Autopilot team and AI team has done an incredible job building real world AI and we’re able to go to wide beta for all full self driving customers in North America this year,”

 

[valem]

Can we post anything anymore on here without having @Knightshade shred it and analyze it line by line?

Or should we shut down this forum until there is no more suffering anywhere in the world?

 

[Knightshade]

There's a bunch of news articles of him saying exactly that: Tesla release FSD beta to North American customers in 2022 – TechCrunch

“The Tesla Autopilot team and AI team has done an incredible job building real world AI and we’re able to go to wide beta for all full self driving customers in North America this year,”

Yeah, as I said that reads to me as the wide release version to everyone in the fleet with FSD.... which will only be "beta" in the same sense all the years-old features like basic AP are, rather than the current narrow development/testing FSDBeta stuff.

But wide release city streets is already years late, and I wouldn't hold my breath this time either.

 

[Daniel in SD]

There's a bunch of news articles of him saying exactly that: Tesla release FSD beta to North American customers in 2022 – TechCrunch

“The Tesla Autopilot team and AI team has done an incredible job building real world AI and we’re able to go to wide beta for all full self driving customers in North America this year,”

Looking forward to @Knightshade arguing next year that Elon only said they were able to do it, not that they would actually do it. Technically they're able to do it right now.

 

[fcastrillo86]

I'm personally holding out hope that FSD goes wide release this year, whether it be in true beta form like it is currently, or maybe similar to how some AP features like Autosteer or Navigate on Autopilot are flagged as beta features.

I get that the goalposts for FSD has continued to move further out every year and it's frustrating. However, Tesla is continuing to grow, and as a result, there are increasing amounts of eyes and attention from the public and investors alike on the company and Elon's ever moving FSD timelines/ambitions. Something has got to give at some point IMO.

 

[Knightshade]

Looking forward to @Knightshade arguing next year that Elon only said they were able to do it, not that they would actually do it. Technically they're able to do it right now.

I mean, by that logic, he also kept his promise of having robotaxis 'approved' in at least one jurisdiction by end of 2020.... since they're approved and legal in like half a dozen US states.

Tesla doesn't have a robotaxi to deploy there (or anywhere)- but they're approved!


As to wide release.... I think it's possible they'd wide release L2 city streets by end of this year. Not super likely, but at least more likely than last year. Like last year I'd have said 1% chance. This year I might go as high as 25-30% based on the current beta, and whenever 11 happens that % could well increase significantly. We'll see.

I think wide release L2 next year, especially if HW4 is easily retrofittable, is a pretty high chance though. Maybe even L3 highway if single stack and HW4 can provide a complete OEDR for a highway ODD.

I think anything beyond that will require more upgrades than just HW4. (more and/or better cameras at least, which may or may not be practical on the current fleet)