FSD 12.3.6- MOST HUMAN DRIVE YET!
ฝัง
- เผยแพร่เมื่อ 18 พ.ค. 2024
- FSD 12.3.6 had its most human-like drive yet, managing turns, stops, pedestrians and speed bumps just as well as most drivers. Watch to see where it can still improve!
0:20 Great roundabout and speed bump reaction
0:55 Awesome oncoming vehicle and speed bump behaviour
2:55 Great merging behaviour
4:48 Dodges oncoming traffic like a pro
5:05 Great safe left turn
6:47 Sensible acceleration
8:40 Lane change to avoid bus
9:20 Too cautious at crosswalks?
10:15 Fantastic pedestrian avoidance
10:45 Final thoughts
FSDBeta 12.3.6 SW: 2024.3.25
Driven in FSD (Spervised) Average setting
Auto Speed Adjust
Filming Gear:
Camera: Insta360 GO 3
Mount: Delkin Devices Fat Gecko Dual Suction Camera Mount (see Amazon)
#FSD #Tesla #FSDBeta #ModelY #fsdsupervised #fullselfdriving - ยานยนต์และพาหนะ
Really enjoyed this drive from FSD. It delivered across the board. Expecting even better results in v 12.4. 😊
Thnks Nice housing in that part of the world!:)
All that you pointed to ….is fixable I would think…
You are right!
To make this experience more realistic, the fsd also needs to keep learning every day. For example, if it didn't notice a pothole on the road and drive hard through it, it needs to memorize this and not enter the pothole on the next pass. If it can even transfer this experience to the vehicles around it, something amazing will happen.
That would be fantastic, but will require next next level compute! When a fleet can speak to each other they could open up a lot of things, but I’m not sure how keen individuals would want to have their vehicle speaking with other vehicles. Even anonymously
They need to do something about dips In residential neighborhoods. I have to take over every time cause it just flies over the dips around 25-30mph
12.4 is coming out now to the chosen few!
Can it recognise potholes? There are everywhere here now days
@@AndrewTSqnope. Not in my experience. I’d be scared to test it out too. Would be interesting to see others’ feedback.
It will only work if ai advances at an exponential pace, I agree that it's far from being actually able to be called fully self driving but I do believe there will be self driving cars in at most 10 years
With the additional investments in computing, I’m expecting the improvements to expedite over previous pace. Fingers crossed.
Most human drive = it bumped a car while parking, was tailgating the car Infront on the highway, never used its blinkers, and was in the wrong sidelines . That is humans to me anyway😂
😂😂😂 maybe it prevents me from being too human!
wooo....the stabilization effect was way too bad, it's swerving left and right all the way, have you tried to tune down the stabilization level?
Will fix that! Got a new camera and trying to learn it! 😂
most human? it will kill then
tru bro
A very easy drive for a human, but FSD has problems:
0:35 Watch the speedometer, FSD didn't slow down much for the speed bump.
1:20 FSD is still going too fast over the speed bumps. 23 mph on this one
3:44 A human driver could have easily went around the truck turning Left
8:55 FSD drives very slowly, and for no reason
9:05 Human must tap the pedal. *Intervention*
9:45 FSD is too slow. Human taps pedal. *Intervention*
FSD still has a long road ahead .
.
Long road ahead...nice pun but I think the road ahead isn't as bumpy.
I don't see a problem with the speed over these types of speed bumps. Speed bump types vary. Some require much slower speeds. I'd be curious to see how it would perform on a road with different types of speed bumps laid out.
As for the going around the truck turning left....he stated that he seen the aggressive setting would have gone around the truck.
FSD crashes more drives than it completes. WHY DO YOU THINK YOU MUST BE LOOKING AHEAD ALL THE TIME
@@e.r.6147 how many times have you driven with FSD enabled?
@@e.r.6147 i haven't seen *any* FSD videos showing FSD crashing. Unless you can link to some examples, I call BS on this claim.
It's always like this - ML fixes few things and breaks another.
That's still milion years away from robotaxi. An hour without intervention is NOTHING, we need 5 million km without intervention (human statistically needs 1 intervention per 100,000 km, AI needs to be at least 10x better). I'm betting on Mobileye and their approach with redundant and independent systems - Vision, Lidar and Radar, with not so end-to-end ML approach which can prevent AI hallucinations.
I’d also like to see more sensor integration as well.
@@FSDwithme Elan says Vision cameras are enough. Meanwhile, FSD is still full of problems after many v12 point releases.
@@DerekDavis213Cry more lmao
@@MrGiggity890 Tesla owners should cry. FSD doesn't work after *years* of promises from Elan Mush.
@@MrGiggity890 FSD is years behind schedule. what a joke.