EVNow
Well-Known Member
Wait … I’ve been out of TMC for a few days. How did we go from FSD to 300baud ?
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD v12.x (end to end AI)
- Thread starter Buckminster
- Start date
-
- Tags
- #fixphantombraking autopilot chill is not chill elon is an ass fix fsd extreme late deceleration fsd does not commit to go fsd does not maintain proper speed fsd hard and late braking fsd has decision wobble fsd jerky driving fsd late deceleration fsd sometimes may not detect train crossings fsd stops too early too quickly fsd stops too far behind line fsd tailgating in city streets
Don't question. Any topic that can help us forget about FSD is welcomed.
Elon how is 12.4 going?
[Elon] it is going swimmingly and 12.4 is pure
nothing can dampen it. I can say with confidence it is flowing smoothly and is in the pipeline to reach everyone. No one will be without 12.4 by the "end of the week".
[Elon] it is going swimmingly and 12.4 is pure
nothing can dampen it. I can say with confidence it is flowing smoothly and is in the pipeline to reach everyone. No one will be without 12.4 by the "end of the week".
Last edited:
sleepydoc
Well-Known Member
We’re just commenting on the rate of the 12.4 rollout!
Back to FSD (12.3.6)
I while ago I commented how in situations where you need to make a left turn followed by an immediate right and there were two turn lanes, the navigation system would indicate that either turn lane was acceptable. I hypothesized that perhaps poor information from the nav system was responsible for poor lane selection by FSD. Someone else commented how FSD always seems to choose the outside turn lane.
Well, lately I’ve noticed more cases where the nav system only highlights the appropriate turn lane option. Not only that, FSD will often chose the incorrect turn lane in spite of this, and there appears to be no pattern as it’s not always the inside or outside lane.
I’m not sure if there was an update to the nav data or central processing but either way, FSD doesn’t appear to be making proper use of it.
From an inside source here is the video Tesla is using to train FSD on hand gestures by Police/road crews.
I feel better about future support for hand gestures now. Should be ready for FSD any day.
I feel better about future support for hand gestures now. Should be ready for FSD any day.
zoomer0056
Active Member
It looks like the main branch people like V12.3.6 more than the test branch people. The main branch people rated 4.9/5 and the test branch 3.9/5. Is it because the test branch people are exposed to too many versions of FSD for too long and FSD improvements have reduced effect? It's like antibiotics cannot kill viruses 
in the hospital?
in the hospital?
Or likely, the test branch people are X employees hoping to make a statement.It looks like the main branch people like V12.3.6 more than the test branch people. The main branch people rated 4.9/5 and the test branch 3.9/5. Is it because the test branch people are exposed to too many versions of FSD for too long and FSD improvements have reduced effect? It's like antibiotics cannot kill viruses
It's worse than that because there are head-scratching regressions and new failure cases.It looks like the main branch people like V12.3.6 more than the test branch people. The main branch people rated 4.9/5 and the test branch 3.9/5. Is it because the test branch people are exposed to too many versions of FSD for too long and FSD improvements have reduced effect? It's like antibiotics cannot kill viruses
Got the opportunity to drive the same route on FSD12.3.6 with each car yesterday.
HW4 Model Y on 24.3.25 and HW3 Model 3 on 24.14.7
What was interesting is two spots on hwy95 both cars made a lurch to the left to try get over the double yellow center line into oncoming traffic. Both times there really was oncoming traffic, so I took over, but did notice that the HW4 car made less of a lurch and and took less effort to stop it. The HW3 car was a real sharp lurch and took a proper grip of the wheel to keep it in line.
For the first time on 12.3.6 it attempted to take a short right turn exit lane off the hwy. I haven't had 12.3.6 do that before
HW4 Model Y on 24.3.25 and HW3 Model 3 on 24.14.7
What was interesting is two spots on hwy95 both cars made a lurch to the left to try get over the double yellow center line into oncoming traffic. Both times there really was oncoming traffic, so I took over, but did notice that the HW4 car made less of a lurch and and took less effort to stop it. The HW3 car was a real sharp lurch and took a proper grip of the wheel to keep it in line.
For the first time on 12.3.6 it attempted to take a short right turn exit lane off the hwy. I haven't had 12.3.6 do that before
Regression.
Mixed with reminiscence of an age when people weren't injured when a computer crashed.
My first programming was on punch cards on an IBM 360 in their ASDD in south San Jose. I was in high school, and joined the Explorer Scouts' first ever co-educational post to learn about PL1. Later CDC 6400 at UCB, still punch cards, Z-80 8" floppies, Interdata 716 (assembly language, regression to paper tape but it had magnetic core memory (non-volatile, rock tour viable) for the Grateful Dead, really.), etc, etc.
So, I'm a boomer and an fsd "tester". I'm hoping we get FSD working before I loose my license due to antiquity.
