Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

FSD Beta 10.69

This site may earn commission on affiliate links.
But 90% succes rate is terrible when driving a car. We would all be dead if we had that.
Why is everyone expecting perfection then getting horribly disappointed with each release that isn’t perfect? This is software development for complex decisions and circumstances.

Tesla is making iterative changes and improvements to their algorithm which is exactly what I expect. The unprotected left was a known issue. Tesla worked on it and made significant improvements yet people complain that it’s only 90%, not 100% accurate. If you can’t deal with that then you don’t understand what beta software is and should bow out.
 
I'm not entirely sure what has changed from Chuck's Memorial Park Drive two months ago (10.12.2) to now aside from the removal of a lot of traffic cones that caused problems in the previous run, and maybe one phantom braking incident? But 10.69 also cut someone off in a roundabout

But I mean these two videos


Almost identical drive time, you can even click to the same timestamp between the two and the car is at roughly the same spot in each. On the final unprotected left, I'd say 10.12.2 was more aggressive and made a bad move but we might see that when they ramp up the aggression of 10.69 with the next two iterations
 
  • Helpful
Reactions: SPACE_EX
Why is everyone expecting perfection then getting horribly disappointed with each release that isn’t perfect? This is software development for complex decisions and circumstances.

Tesla is making iterative changes and improvements to their algorithm which is exactly what I expect. The unprotected left was a known issue. Tesla worked on it and made significant improvements yet people complain that it’s only 90%, not 100% accurate. If you can’t deal with that then you don’t understand what beta software is and should bow out.
Elon has been - disputedly - quoted as saying 100% was the quantitative goal for Chuck's UPL. See this exchange up-thread, or the shareholder meeting YouTube around the 8:30 mark. It came from one of his comments at the meeting, where he said, amidst general laughter "...we're gonna solve Chuck's turn - yes - yes - 100% - absolutely..." and then saying apparently to Chuck, "... we're at about 90% success rate with your turn - almost 100..."

So it's not hard, if it serves your position, to pin a claim of 100% success hard goal on Elon, with subsequent judgment of failure when it's not 100%. I don't think anyone seriously thought it would be 100%, including Chuck himself who has made it clear that he's not expecting that. I think it's reasonable to interpret the first "100%" comment referenced above as "for sure we're gonna solve the turn", but not that it'll be perfectly olved.

Another justification (again not mine) for attacking the imperfect result is that, indeed, it has to get really, really close to 100% (99.nininini%) as a released L4+ feature. In this context, Elon's second mention of "100" is a simple acknowledgment of that eventual target. But it's daily fodder for ridicule of FSD, that it's not presently within the march-of-nines asymptote in many areas.

Chuck's videos, as ever, are already leading the way with the correct attitude, which is that he's thrilled about the progress but immediately describes the shortcomings and/or uncertainties about what the car is thinking. Almost as if he understands that developing a solution to a complex problem doesn't happen overnight.

While the turn is fun to discuss , this release has a ton of improvements as laid out in the release notes. There are more impressive and less impressive releases along the way, and 10.69 is certainly looking like the former.
 
It would suck if the only feature I am allowed to use here never benefits from the FSD development. Seems like the old stack is not updated in months / years.
[speculation] My theory is eventually the standard AP stack will be replaced with the highway FSD Beta stack. But first we need V11 stack integration and then it will have to iron out regressions and lots of testing and fine tuning. Legacy AP stack may be retired/replaced in about a year. [/speculation]
 
  • Added control for arbitrary low-speed moving volumes from Occupancy Network. This also enables finer control for more precise object shapes that cannot be easily represented by a cuboid primitive. This required predicting velocity at every 3D voxel. We may now control for slow-moving UFOs.
Looks like there's new visualizations when the occupancy network detects things that aren't predicted with labeled cuboids resulting in grayish regions and "Changing lanes away from path blockage" message:

occupancy path blockage.jpg

 
[speculation] My theory is eventually the standard AP stack will be replaced with the highway FSD Beta stack. But first we need V11 stack integration and then it will have to iron out regressions and lots of testing and fine tuning. Legacy AP stack may be retired/replaced in about a year. [/speculation]
A year sounds about right as speculation for the main stable-release version of a replacement highway stack.

If, however, v11 stack integration occurs this year, and then if Elon's prediction of FSD Beta becoming available (initially as an opt-in) to all paid-FSD customers - two big ifs - then effectively we will have earlier availability of the merged-stack NoA to anyone who wants it, while keeping the old-stack NoA for those who prefer it.
 
  • Like
Reactions: pilotSteve
[speculation] My theory is eventually the standard AP stack will be replaced with the highway FSD Beta stack. But first we need V11 stack integration and then it will have to iron out regressions and lots of testing and fine tuning. Legacy AP stack may be retired/replaced in about a year. [/speculation]
Yeah I think you are right - a least a year out. This takes time.
Sucks to be in Europe right now but Elon did say that canada and Norway were next. How he would make it work juridically in Norway , I have no idea about ...
 
Why is everyone expecting perfection then getting horribly disappointed with each release that isn’t perfect? This is software development for complex decisions and circumstances.

Tesla is making iterative changes and improvements to their algorithm which is exactly what I expect. The unprotected left was a known issue. Tesla worked on it and made significant improvements yet people complain that it’s only 90%, not 100% accurate. If you can’t deal with that then you don’t understand what beta software is and should bow out.
And what is “perfection”?

My takeaway from watching the first couple of videos is that it works great. The comments on the video are all like “ a little slow, a little jerky, a little close, a little…” I bet if my wife made a video of how I drive she could have made the same responses as could I if she would have driven.
If you expect a carbon copy of how you would have driven a route than you will be in for a disappointment. The main objective is that it drives safely from A to B and I think Tesla is on the verge of accomplishing that.
 
Capricious lane changes have been my single biggest complaint. I really hope they've sorted that out, or that we get an option to check in autopilot setup that says- "no unnecessary lane changes"
Yep. The vast majority of my interventions aren’t due to an imminent collision or anything dangerous; they’re just to keep the car from changing lanes in the wrong direction with a turn upcoming in less than a quarter mile, sometimes within a few hundred feet. Turn signal comes on to change lanes, i nudge the stalk in the other direction to cancel it. Sometimes, if it is changing lanes from far enough away, I get to play a fun dance where I no doubt confuse the driver behind me as my attempt to cancel the lane change engages the signal too early in the other direction, so I try to cancel that, which activates the signal the other way, and the dude behind me is wondering why the crazy person in front of him can’t pick a direction to go.
 
Chuck finds a serious issue on 10.69. I suspect it is because there is no center line. I know FSD Beta tries to center itself when on residential streets like this that don't have a center line. So I think it is trying to center itself despite there being a truck in the path. Definitely odd and needs to be fixed.

 
  • Informative
Reactions: Thp3
This is nothing new, he has been reporting it for many versions.
Yeah nothing new. The additional issue shown is the extremely long time it takes to get into position. This is something it should be capable of doing in 10.69 quickly.

Push all the sliders to 11! (Though we get a sense for why they may not already be at 11 with the example just above.)