I'm gonna wait for 2.5 before deciding if Tesla has a handle on this. "Once is happenstance, twice is coincidence, three times is a pattern."
-
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 jerky driving fsd late deceleration fsd stops too early too quickly fsd stops too far behind line fsd tailgating in city streets
Say what????? Supervised is just an added word to help placate the systems acceptance and promote paying attention to the newbies. There is absolutely NO such thing as an UNsupervised L2 system.
That is what he is saying. "Supervised FSD" is a Level 2 system, and always will be. The prediction is that they will add a Level 4 Robotaxi/"Unsupervised FSD" option in the future, and that it will cost more. (Though original FSD buyers did essentially buy that.)
And I predict I will win the lotto.
Are cars are great and developing an excellent L2 (and I'm happy) but L4 is not likely in the realm of possibility with the hardware limitations. I bet on 8/8 (if Tesla covers it) we will find the Robotaxi will have HW5, MORE camera coverage (4 cross traffic and front bumper at least) at higher resolution/bit depth. Also wouldn't be surprised if they also include LiDAR and/or HD Radar. Serious Robotaxies require serious hardware.
Last edited:
gottagofast
2022 Model 3 Performance
Yea this means people that paid 15k for FSD will essentially be downgraded lol
It can be looked at 2 ways.
- FSD(S) is out of Beta and a delivered system now, so a done deal as far as FSD goes
- Elon/Tesla can plausibly say they are working (indefinitely) towards getting current cars to L4 (just like it has been)
PianoAl
Active Member
What improvements are you looking forward to? Mine behaves fine on the freeway.
One improvement is the switching of stacks can be awkward and case hiccups. Also this would free up a lot of space eliminating V11 code. We could maybe get some more fart sounds.
PianoAl
Active Member
I gave up on Autopark yesterday. It took so long and had so many back-and-forth corrections that I wanted to get out and announce to any onlookers that I wasn't driving.
old pilot
Member
And I predict I will win the lotto.
Tesla Surprisingly Becomes Luminar's Largest Lidar Customer, Contradicting Musk's Past Criticism
Tesla Inc. (NASDAQ:TSLA) CEO Elon Musk has described lidar sensors as a “crutch” for autonomous vehicles. Yet Tesla has become the largest customer of lidar manufacturer Luminar Technologies, Inc. (NASDAQ:LAZR). As disclosed in Luminar’s earnings report on Tuesday, in the first quarter of 2024...
willow_hiller
Well-Known Member
Given how long these models seem to take to train, I wouldn't be surprised if 12.3 was the last model that was still trained prior to Tesla's massive compute acquisition. If they're really no longer compute-constrained, they could even train one new model per chill/average/assertive mode.
I went back to my post from a while back (v12.3.3 at that time) to answer this question. I'm hoping the new AI stack will address some or all of these problems on the freeway.
I have not had a single incident of disengagement on highways yet after 11.3.9, and FSDs. The disengagement on city roads have been due to ambiguous construction zone cone placements.
Last edited:
mgs333
Member
On my wife's newer 2023 Model 3 without USS, the autopark worked fine, actually usable. On my 2018 Model 3 with USS, the autopark has been absolutely terrible like you said.
I'd set up the car perfectly so that all it has to do is back in at the angle I left it at, then the steering wheel turns the wrong way, goes in at a different angle then has to start going back and forward multiple times.
mgs333
Member
WTF does fixed even mean? The nags are back in 12.4? Or much reduced or gone in 12.4?
Similar threads
