-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
haha same ole redefining and defending what Elon says to explain what he meant once things doesn't happen as he hyped it up to be.
My mind is BLOWN that tesla fans never change.
Elon also said "Beta 9 addresses most known issues" only two days ago- I'd think killing your users at every unprotected left would be a known problem that most people would assume they would focus on fixing during a foundational rewrite and at a company where safety is top priority.
There's a Elon tweet for every point of view as long as you ignore all the other tweets!
What I think is most interesting is how minimal the changes are for supposedly a complete rewrite. I have a feeling what he really meant was that some corner was rewritten- like distance estimation to remove radar But the rest of the logic is the same as before, so basically the same issues continue to exist. In my skeptical view, this is just another place where Elon is disingenuous in his communications, hyping small changes with imprecise language, knowing people will assume it means much more.
Remember, this is 8+ months of progress since V8.. If this is the rate of improvement, we're a very long ways away from generalized self driving.
Last edited:
Daniel in SD
(supervised)
Those are older tweets. He seems a little less optimistic/delusional lately. haha.
Also, you can see from my posts that I'm not a fanboy.
And 6 years since the self driving was solved.
Anyway this once again shows, that you should not give any credit to Musk’s promises about FSD.
Daniel in SD
(supervised)
Ok. I'm wrong. This is pretty indefensible. haha. I guess Elon will never change.
Unless they did zero testing before release this can't possibly be true. Though maybe it depends what they mean by "address", he didn't say fixes most known issues...
Nice. But I need
Terrible FSD operator. He should spend less time screaming about how hype he is and more time actually intervening when the car is about to cut someone off or otherwise screw up. Take over and report when the car is clearly confused, don’t let it fumble it’s way through while there are other drivers waiting on you. Tesla monitors the disengagements so let them take a look and fix instead of inconveniencing other people for your own fun.
powertoold
Active Member
I'll admit I was wrong about this V9 release. Yes, it is very impressive and better than 8.2, but it's not like Tesla spent the last 5 months actually fixing all the issues. I think they spent the last 5 months on making vision-only at least on parity with radar, which is an amazing feat in itself, but it didn't result in dramatically better driving performance.
I think the main path forward for Tesla will be to create their own cloud maps that gives the car a heads up about what lane to be in (kinda like upcoming traffic controls).
Otherwise, if we subtract the lane choice issues, I think V9 is very impressive and still on track for human level+ FSD (which is my personal threshold for wide release), although I think it's still at least 3+ months away.
I think the main path forward for Tesla will be to create their own cloud maps that gives the car a heads up about what lane to be in (kinda like upcoming traffic controls).
Otherwise, if we subtract the lane choice issues, I think V9 is very impressive and still on track for human level+ FSD (which is my personal threshold for wide release), although I think it's still at least 3+ months away.
mikes_fsd
Banned
That seems problematic.
The goal has been the same since they made it -- the goal -- public in October 2016:
That is the goal.
So, because you do not understand something, we are supposed to abandon it for something you also do not understand (simulation)?
Then it will never be done, because you have to have the performance data of the current feature set to compare your improvement against.
Based on one single video of a handful attempts of the same left turn? Am I following your logic?
powertoold
Active Member
Turns and overall confidence / assertiveness have improved a lot. It does extremely well when there's only 1 lane of travel. When there are multiple lanes, it still struggles with lane choice.
It also seems the overall refresh rate has improved 50-100%. That's why the wheel is much smoother on turns.
Daniel in SD
(supervised)
So, robotaxis. I guess what I struggle with is why release beta software with so many known bugs? It seems like you're going to get so many bug reports of bugs that you already know about that it will just be a waste of time.
I'm saying that if you can reproduce the bug in simulation then why test it in the real world? Get it working in simulation where you can test a million different variations very quickly, get it working, and then test it in the real world. The knock against simulation has always been about testing edge cases, not basic functionality like turning left.
I'm talking about release to the public, not testing by trained test drivers.
I watched a couple other videos too. It looks like they could have easily seen these bugs. I'm skeptical that they simulated all his previous failed left turns and verified that they worked on this build.
Someone, please put some massive white boxes in the road, at blind corners, make fake bollards, fast moving tall boxes dressed with 'clothing' and let the car go at them. We NEED to know whether it will hit them or not. IS it going to hit the monorail pillars? It's extremely important to know the answer. Avoiding them manually tells us nothing.
Again, test with BOXES or BALLOONS please, of course don't put your car at any risk.
diplomat33
Average guy who loves autonomous vehicles
I noticed several videos now where the path planning jumps around a lot when stopped at an intersection.
Here is one example where you can see the path planning is all zig zaggy to the left when the car needs to turn right. The path planning keeps jumping back and forth a lot. It is like the planning is not sure what to do.
Here is source video:
Here is one example where you can see the path planning is all zig zaggy to the left when the car needs to turn right. The path planning keeps jumping back and forth a lot. It is like the planning is not sure what to do.
Here is source video:
powertoold
Active Member
I noticed several videos now where the path planning jumps around a lot when stopped at an intersection.
Here is one example where you can see the path planning is all zig zaggy to the left when the car needs to turn right. The path planning keeps jumping back and forth a lot. It is like the planning is not sure what to do.
Here is source video:
I notice that as well, but when it comes time for it to go, the path stabilizes. It was also the case in that video, but I'm not sure why Chuck intervened because the path was correct once the car moved up a few feet, probably just out of caution.
diplomat33
Average guy who loves autonomous vehicles
Personally, I do not think Tesla is 3+ months away from "human level+ FSD". But the thing is that without real data like accidents per million miles, near misses per million miles or interventions per million miles, etc... it is impossible to really quantify how close or far Tesla might to be "human level+ FSD". So we are all just guessing wildly based on what videos we decide to cherry pick.
powertoold
Active Member
I agree, but I really do think we'll know it when we see it in videos. Right now, it's making many unnatural choices. It's difficult for me to see Tesla doing a wide release until the car acts naturally most of the time. Otherwise, most people won't even use city streets fsd (even some fsd beta testers admit that they don't use it much, other than to show it off), although it's possible Tesla won't place too much emphasis on comfort / practicality for the initial wide release.
diplomat33
Average guy who loves autonomous vehicles
I wonder if it is a bug where the software is showing all possible path predictions the car is considering. If you look at the paths closely, it seems to be jumping between different possible paths (go straight, turn left, turn right). And then when the car starts to move, it settles on the actual path it has chosen.
Similar threads
