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

FSD Beta Videos (and questions for FSD Beta drivers)

This site may earn commission on affiliate links.
Do you use AP in the city now ? Took me a while to start using it - but now I use it all the time.

1631823286619.png


Back to my question- people are already using current AP on city streets all over the place, in direct contradiction to Tesla's own manuals. The manual we point to when someone "relied" on AP and it crashed, and we point out that in no way, shape, or form could this have been Tesla's fault, as it's an L2 system.

How could releasing city suburban streets autosteer be worse for safety than what they have now? Why is Tesla waiting?
 
Do you use AP in the city now ? Took me a while to start using it - but now I use it all the time.

On occasion I try it, but it's essentially completely useless to me. As far as I can tell, I'm much more engaged in the driving task and safer if I am not using it on city streets. It's very unpleasant to have to be confirming traffic lights all the time, etc., predicting when it is going to veer out of lane into a turn lane, etc. Not helpful in current form, for sure, for me. (Certainly regular EAP was no worse as far as I can tell.)
 
Wouldn't Tesla want to know when the driver is manually spinning the speed up or down so they can determine why? For example is it a mapping error or did FSD incorrectly read a speed limit sign? Immediately dismissing the operator feels a bit harsh don't you think since they are providing feedback to Tesla that could improve FSD?
I really doubt Tesla is looking at every time the driver changes the TAAC speed. But sure, they could get some interesting data from that.

The issue now is that we have a bunch of YouTubers going "Look at how AWESOME FSD 10 IS!!!!!" when the reality is that they are gaming it, manually setting speeds constantly in order to get it to look like it succeeded, when in reality it will fail without a lot of intervention. It's just intervention that is a lot less obvious to the casual observer than grabbing the wheel. The YouTubers testers that press the throttle with their foot without saying anything are the same. These are not cases where it mis-read a speed limit sign. The screen clearly shows the correct limit of 25 MPH while the tester has it set at 8 MPH.

I'm not saying we dismiss the operator. I'm saying we dismiss these videos as any kind of evidence of how good FSD is or how it has advanced since the last release. V9 managed to not hit a monorail pole with TACC at 15 MPH once. Now V10 did it 3 times while set at 8 MPH. PROGRESS!
 
Last edited:
I really doubt Tesla is looking at every time the driver changes the TAAC speed. But sure, they could get some interesting data from that.
The issue now is that we have a bunch of YouTubers going "Look at how AWESOME FSD 10 IS!!!!!" when the reality is that they are gaming it, manually setting speeds constantly in order to get it to look like it succeeded, when in reality it will fail without a lot of intervention. It's just intervention that is a lot less obvious to the casual observer than grabbing the wheel. The YouTubers testers that press the throttle with their foot without saying anything are the same.

I'm not saying we dismiss the operator. I'm saying we dismiss these videos as any kind of evidence of how good FSD is or how it has advanced since the last release. V9 managed to not hit a monorail pole with TACC at 15 MPH once. Now V10 did it 3 times while set at 8 MPH. PROGRESS!
Some testers take the car on routes where it has struggled before, difficult street-crossing scenarios. Frenchie and Chuck are two. Often they end up annoyed at its failure.

It's possible that some testers avoid difficult scenarios where they know it would do badly. That's another way to game the test.
 
  • Like
Reactions: gearchruncher
View attachment 710205

Back to my question- people are already using current AP on city streets all over the place, in direct contradiction to Tesla's own manuals. The manual we point to when someone "relied" on AP and it crashed, and we point out that in no way, shape, or form could this have been Tesla's fault, as it's an L2 system.

How could releasing city suburban streets autosteer be worse for safety than what they have now? Why is Tesla waiting?
On a previous post of yours, I believe you asked if I use AP on city streets, or not on highways/freeways, I can't remember. It doesn't really matter.
My answer is: I do, I use it pretty much as often as I can. At this point I feel it's synergic, you know, work commute, same country roads to usual places you go etc. and I am fully aware that this is my decision and my responsibility. I would never blame Tesla for an accident. I successfully anticipate 99%+ of the car's decisions and for the 0.99999...% I react well within time - not to avoid an accident, but to stay on track.

And then there's that 0.000001%. I did have an accident when I hit a flat cone (bottom pointed towards me and not reflective, at night). I called no one but my insurance company about that. Could Tesla have avoided that? Absolutely. Could I? Absolutely. The blame is either on me or whatever entity is responsible for safe roads in here. Not Tesla. The same applies for any decisions I make while driving (or monitoring 😁) my car.
 
Gotta "love" hyperchange fanboying over bad behavior like it is a huge success:


This is not a success. This is terrible behavior for any AV. FSD Beta makes an illegal right turn from a straight only lane, wobbles as it tries to make the turn, almost hits the car on the right, turns too wide towards the parked cars, and finally straightens out in the lane.
 
Last edited:
This is not a success. This is terrible behavior for any AV. FSD Beta make an illegal right turn from a straight only lane, wobbles as it tries to make the turn, almost hits the car on the right, turns too wide towards the parked cars, and finally straightens out in the lane.
With the speed limit set to 10 MPH!
No, see this is very impressive, Tesla is clearly right on the cusp of solving generalized AI. Just don't use it in city centers, those are edge cases.

If we all had this "beta" and someone posted a video of it being used this way, people would be yelling at them to take over and not endanger the public with these stunts. But because we don't get to use it ourselves, we give a pass because we're so thirsty for content.

Elon is a marketing genius, through and through, and I fall for it as much as anyone.
 
Last edited:
Gotta "love" hyperchange fanboying over bad behavior like it is a huge success:


This is not a success. This is terrible behavior for any AV. FSD Beta makes an illegal right turn from a straight only lane, wobbles as it tries to make the turn, almost hits the car on the right, turns too wide towards the parked cars, and finally straightens out in the lane.

That whole video was pretty embarrassing. I'm not sure if I've ever seen a tester let it fail so badly in a populated area. I doubt it would make the same mistakes had there been no traffic or pedestrians around, so I'm not against testing such scenarios*... but he let the car cross the line too many times into reckless territory.

\* I mean this as a preemptive rebuttal to people who say FSD needs to go back to closed tracks without traffic or pedestrians. I don't agree with that because it's exactly the presence of other actors on the road that needs to be tested and worked out at this stage.

But this testing only works when the testers disengage before allowing the car to do something unsafe. This guy really makes the whole program look bad.
 
Gotta "love" hyperchange fanboying over bad behavior like it is a huge success:


This is not a success. This is terrible behavior for any AV. FSD Beta make an illegal right turn from a straight only lane, wobbles as it tries to make the turn, almost hits the car on the right, turns too wide towards the parked cars, and finally straightens out in the lane.
@gearchruncher, this is probably why they're not doing wide release. As far as I can tell people who are using FSD Beta drive far worse than they would normally. Everyone on the road is unwittingly a beta tester too.
 
#FSDBeta v10 Unprotected Left Turns - Better Late than Never!

Compared to Beta 9, Beta 10 is doing much better in that it realizes there's a left turn destination across the grassy raised median. It's not quite directly comparable with the smaller visualization (@Chazman92 my guess is you need to shrink the navigation directions by tapping the top area), but here's Beta 9 visualizing a path into the 3rd lane for the wrong direction and usually ended up turning right as it couldn't predict a reasonable path to the left:

left 9.jpg


Whereas Beta 10 actually has a path to the left past the median:
left 10.jpg


I would guess one of the main "rewrites" for Beta 10 is the deploying of the Video Module for intersection and lanes predictions as presented at AI Day. (Whereas Beta 9 introduced the Video Module but primarily for predicting velocity/position as part of radar removal.)
 
Funny you say that. I love using AP in heavy traffic, in my experience it makes a wonderful job with stop-and-go. Never had an issue, not the tiniest one in this scenario. What makes you not like it in heavy traffic? Or is my interpretation of heavy traffic different? I presume it could also mean "busy", but not "stop-and-go". Is that it?
Heavy city traffic (both vehicles and foot).
 
  • Like
Reactions: helvio
View attachment 710205

Back to my question- people are already using current AP on city streets all over the place, in direct contradiction to Tesla's own manuals. The manual we point to when someone "relied" on AP and it crashed, and we point out that in no way, shape, or form could this have been Tesla's fault, as it's an L2 system.

How could releasing city suburban streets autosteer be worse for safety than what they have now? Why is Tesla waiting?
That bit hasn't changed in a long time - even though Tesla keeps releasing city specific features - and probably even recognized some revenue based on that. We have also discussed this issue ad nauseum.
 
this is probably why they're not doing wide release. As far as I can tell people who are using FSD Beta drive far worse than they would normally. Everyone on the road is unwittingly a beta tester too.
When I commented yesterday on how bad this was, many people told me that it was quite impressive, and shows how amazing FSD is. I was asking those people why it couldn't be released if a video like that one is great.

But I get crickets or jokes on why they aren't releasing an amazing, well performing feature.
 
When I commented yesterday on how bad this was, many people told me that it was quite impressive, and shows how amazing FSD is. I was asking those people why it couldn't be released if a video like that one is great.

But I get crickets or jokes on why they aren't releasing an amazing, well performing feature.
They aren’t releasing because it isn’t ready, but the videos do show impressive improvement
 
  • Like
Reactions: ahoen117 and Mengy
Gotta "love" hyperchange fanboying over bad behavior like it is a huge success:


This is not a success. This is terrible behavior for any AV. FSD Beta makes an illegal right turn from a straight only lane, wobbles as it tries to make the turn, almost hits the car on the right, turns too wide towards the parked cars, and finally straightens out in the lane.
I said it earlier in the thread, Tesla needs to take the Beta way from him if he keeps doing this. He’s going to get into an accident. The dude is too busy saying “This is dope” to actually monitor the car.