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

Elon: FSD Beta tweets

This site may earn commission on affiliate links.
for those who appear to lack critical thinking/reading skills: (you know who you are).
Missy's role is NOT to be head of NHTSA.

Go back and carefully read the reputable articles from reputable sources to find out what role she has been assigned to fill. Note: Facebook isnt considered a reputable news source.
Thanks. I didn’t research it myself but I thought it was for head of NHTSA.
 

“Most likely” FSD beta 10.3 coming Friday (night). So could be on time. Not clear that there will be significant feature changes in 10.3.


And visualizations come to all vehicles “when we’re out of beta.”

It’s left ambiguous whether that means when we get to FSD Beta (rather than the beta of FSD Beta) or to FSD. I suspect the former.
I don't think 10.x will see any feature changes going forward except that it will fail less often at what it tries to do now. 11.x is the next chance for feature changes. What do you classify as a feature change?
 

And visualizations come to all vehicles “when we’re out of beta.”

It’s left ambiguous whether that means when we get to FSD Beta (rather than the beta of FSD Beta) or to FSD. I suspect the former.

Interesting that Elon wants to wait until they are out of beta just to release UI. I seem to recall that they gave that "FSD sneak peak" where users got updated visualization of traffic lights before the feature came out. You would imagine that they could give the fancy FSD UI even to non-FSD users. Maybe the UI is tied to the features and can't easily be displayed for non-FSD? It is possible that Elon wants to wait until the FSD visualizations are more polished before letting non-FSD users have it.

Considering that some features are still beta years later (Isn't autosteer still labeled as "beta"?), I doubt that Elon means when FSD itself is not beta anymore. He probably means when FSD Beta is released wide and not considered "early access" anymore. Just my guess.
 
I don't think 10.x will see any feature changes going forward except that it will fail less often at what it tries to do now. 11.x is the next chance for feature changes
At least the recent FSD Beta "major" versions have been for significant technical changes but the overall "feature complete" FSD is "just" doing things better.

FSD Beta 9 introduced Vision-only stack resulting in better predictions of other vehicle attributes, e.g., velocity and position.
FSD Beta 10 introduced Video/Feature queue needed for the Lanes network to predict which lanes to get in-to/out-of instead of relying on map data.
FSD Beta 11 will likely introduce single stack, which should improve city street situations that look more like parking lots or highways.

There have been various improvements in all FSD Beta versions that are more noticeable, e.g., visualization improvements and clear-to-go boost, and some are purely under the hood to collect more data via shadow mode likely to evaluate upcoming major version changes.

I wonder if the 10.3 technical details / release notes will be to better direct FSD Beta testing or just that there's a larger audience that information will be relevant to:
 
  • Like
Reactions: qdeathstar
FSD Beta 11 will likely introduce single stack, which should improve city street situations that look more like parking lots or highways.

Not sure what city street situations look like parking lots or highways. Those are very different driving environments.

I think single stack will smooth out the transition between city driving and highway driving. It will mean that we can go from A to B on the same FSD stack instead of switching between stacks. It may also bring some improvements to highway driving as well.
 
  • Like
Reactions: jebinc
Yes - so theoretically improvements in highway driving. But how will it affect city fsd (including parking lots !) in anyway ?

Single stack = changing freeway NOA to be more like City FSD.
I think it's tightly related to Summon. Today the images flicker too much and that's transported to the steering wheel adjustments. If camera fusion is used for Summon (which is currently only used in Parking Lots/private drives), it should stop flickering.

Another possibility is vision-based perpendicular parking.
 
I think it's tightly related to Summon. Today the images flicker too much and that's transported to the steering wheel adjustments. If camera fusion is used for Summon (which is currently only used in Parking Lots/private drives), it should stop flickering.

Another possibility is vision-based perpendicular parking.
Yes - makes sense to include summon / parking lots in city FSD.

But the basic question remains - do you want to fix pressing city FSD issues (like FCW / sudden slow downs, very hesitant round-abouts, jerky steering wheel action) or spend those resources adding features to the city fsd stack ? I know where I'd spend the resources.
 
Yes - makes sense to include summon / parking lots in city FSD.

But the basic question remains - do you want to fix pressing city FSD issues (like FCW / sudden slow downs, very hesitant round-abouts, jerky steering wheel action) or spend those resources adding features to the city fsd stack ? I know where I'd spend the resources.
Makes sense. But sometimes improvements from one can benefit the other, bringing that Summon topic, which is currently a joke, to a low hanging fruit. I wouldn't be opposed to making that better because at some point it has to be. At the same time they might be collecting more feedback to improve other functions, so it's a good "in between" effort. I wouldn't look at those issues as mutually exclusive.
 
  • Like
Reactions: edseloh
Makes sense. But sometimes improvements from one can benefit the other, bringing that Summon topic, which is currently a joke, to a low hanging fruit. I wouldn't be opposed to making that better because at some point it has to be. At the same time they might be collecting more feedback to improve other functions, so it's a good "in between" effort. I wouldn't look at those issues as mutually exclusive.
Yes, if they have extra resources. Otherwise I'd concentrate on top 5 or 10 issues fsd testers are reporting on.

Its possible Tesla has a separate set of devs working on single stack (esp. the NN side, if most of the bugs are on the planner side). But I'd never prioritize this over fixing critical bugs in city fsd. Fixing bugs in city FSD would help reduce chances of future accidents which could damage the long term plans of Tesla.
 
  • Like
Reactions: edseloh and helvio
Not sure what city street situations look like parking lots
I include large private plaza/complexes that have their own service roads that connect the parking lot aisles, but in general, these parking lots can look very much like regular city streets, but they're much more likely to have odd features such as sudden bends in the road, sharp turns and non-straight curbs/medians. Here's a simple example where FSD Beta 10 could potentially be improved with a single stack trained on data from parking lots:
parking lot.jpg


With FSD Beta 10.2, I've encountered normal city streets where it failed to stay within the lane because the whole road suddenly shifts to the left. I've also had situations where it gives up on making a sharp left turn and instead makes the easy right turn because the predicted path to the left drops below some threshold. I've also had to disengage for a very short curb extension (small platform for pedestrians to wait) because the visualization predicted a straight curb.

Yes, there are additional complexities to parking lots such as the very close intersections for each aisle, but that's also not too different from a regular street where FSD Beta gets confused by a parking lot entrance/exit that is right next to the actual street you want to turn to.
 
  • Like
  • Helpful
Reactions: tmoz and diplomat33