-
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 doesn't read signs properly 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
KArnold
Active Member
Think optimisticly.
"At this point, v12 works amazingly with a wide release in 2 weeks!"
Wonder if ChatGPT could create a deep fake of Elon saying this. Or we could just wait for that tweet, which really means 6+ weeks from then.
"At this point, v12 works amazingly with a wide release in 2 weeks!"
Wonder if ChatGPT could create a deep fake of Elon saying this. Or we could just wait for that tweet, which really means 6+ weeks from then.
OxBrew
Active Member
Could background image lower probabilities for an action stack up from training and become higher probabilities when high probability objects are not present?
For example, an image with a slowing leading car (car got bigger in next frame) has a 90% correlation for why human driver (training) slowed down. Around the car, hundreds of other pixel combinations get labeled with 0-10% probability for why human slowed down.
Cruising along on an open road, theres a constant stream of 0-10% (IDK, some low threshold number) pixel combinations flashing all over the images. Could there by chance be some random repeating pixel combination that coincidentally occurred along with a 90% event (actual lead car slow down) that ends up above a trigger threshold, but is masked by the 90% event? Remove the 90% event, and some of that background noise ends up looking like a 40% event, and the AI driver slows a bit just in case? Phantom slowing, or all and sundry undesirable behavior?
Apologies for the inexpert language and understanding, just a thought I had. No idea if anything like this is really going on under the hood.
For example, an image with a slowing leading car (car got bigger in next frame) has a 90% correlation for why human driver (training) slowed down. Around the car, hundreds of other pixel combinations get labeled with 0-10% probability for why human slowed down.
Cruising along on an open road, theres a constant stream of 0-10% (IDK, some low threshold number) pixel combinations flashing all over the images. Could there by chance be some random repeating pixel combination that coincidentally occurred along with a 90% event (actual lead car slow down) that ends up above a trigger threshold, but is masked by the 90% event? Remove the 90% event, and some of that background noise ends up looking like a 40% event, and the AI driver slows a bit just in case? Phantom slowing, or all and sundry undesirable behavior?
Apologies for the inexpert language and understanding, just a thought I had. No idea if anything like this is really going on under the hood.
zoomer0056
Active Member
Well, I had to...
ChatGPT doesn't know about the "six months for sure" part of that statement.
Last edited:
AlanSubie4Life
Efficiency Obsessed Member
And would obviously be nowhere near good enough for the driving task, even if it could operate in real-time, with current technology.
It’s over, folks. For this hardware (3/4).
Glad to see our resident "vacillator-in-chief" has flipped back to “insurmountable obstacles” viewpoint. All is right with the world.
Who could have predicted this?
I am still very optimistic it will be a better driver’s aid than v11. Just have to keep alert. I doubt it will be very useful. Just better.
I'm a little concerned about the ability of people to monitor this. Not sure what all these YouTubers are doing. Just disengage way before anything bad happens. And then just tell people you disengaged because it was going to screw up, though it had not technically done so yet.
Last edited:
Sounds absolutely nothing like Elon lol
AlanSubie4Life
Efficiency Obsessed Member
sleepydoc
Well-Known Member
No, I actually saw a vampire braking for a truck of garlic.
Bets on how many people now think Tesla will definitely make a jump to 2024.X not understanding TeslaScope said "may"?
on 12.3 by this weekend. Plenty of microseconds left. 
EVNow
Well-Known Member
Yes - the "ChatGPT" moment people have been waiting for is not good enough. Chat GPT can make 1/10 bad outputs and people only look at the 9 good ones. With FSD everyone looks at that 10th bad one. For reliable driving FSD should be making like a mistake once in a million (or more).
Anyone want to leak what Teslascope shared in their subscriber only post? I'm not going to pay for rumors, as curious I am.
How many crashes have there been with 12.x so far? Is it just the parking lot one? I wonder if Tesla will put more guardrails / restrictions on 12.x until it's better trained for parking lots / Smart Summon? After seeing the various failures of 12.2.1 even outside of parking lots, it's not that surprising Tesla has held off on expanding the rollout to reduce risk of more safety critical issues.
Hopefully Tesla's metrics on 12.x safety related disengagements are not shadowed by general disengagements of 11.x where the latter probably has a lot more comfort/convenience type disengagements, so the disengagement rate of 12.x is probably lower but of the disengagements that do happen, there might have been an increase in safety issues.
