I'm guessing Tesla wanted to get some 12.x deployed to some non-employee customers, and 12.2 wasn't going to be ready to fix known issues of 12.1 before Wednesday's Q4 and FY 2023 Update. One way to still achieve that is to increase the strictness of nags and thresholds for heavy rain / poor weather to deactivate 12.1.2:“They indicated the nag has been dialed up to ensure drivers are paying extra attention.”
-
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
AlanSubie4Life
Efficiency Obsessed Member
It did seem super slow and did quite poorly here (note how far behind the lead traffic it fell), as was noted by Whole Mars. Didn’t look at the real time to figure out why it did so badly. Just somehow waited way too long.
It's been configured to be conservative while they betatest it in small scale. When they feel more confident in the system, they will allow it to drive faster by editing a few variables.
I noticed the blue driving path visuals are extremely short on that V12 build. It doesn't show it's going to make a right turn until it's partway through it.
AlanSubie4Life
Efficiency Obsessed Member
Just was contrasting this with the distinct assertive behavior that I noted it shows elsewhere, which seems to exceed anything we have seen in current v11 builds. (Not necessarily bad - depends on how it behaves in a situation which has mostly hidden hazards, of course.)
Being assertive and fast is potentially very good. But it’s complicated. And dialing to conservative doesn’t necessarily translate to safer - it can make it less safe.
I’m glad someone else is a believer in the slider theory though.
It’s very controversial.
Last edited:
First highway test of FSD 12 in very heavy rain was not impressive speed limited to 60 and kept getting red hands forced disengagement multiple timesJust was contrasting this with the distinct assertive behavior that I noted it shows elsewhere, which seems to exceed anything we have seen in current v11 builds. (Not necessarily bad - depends on how it behaves in a situation which has mostly hidden hazards, of course.)
Being assertive and fast is potentially very good. But it’s complicated. And dialing to conservative doesn’t necessarily translate to safer - it can make it less safe.
I’m glad someone else is a believer in the slider theory though.
You think this is the end2end? Or does it sound more like heuristic code aka "for this version, if heavy rain -> limit speed= 60mph & force hand on wheel"...
powertoold
Active Member
Wow, V12 is insane, my socks have been blown off, the first drive was phenomenal, a refined V12 would be endgame FSD
We still need to see if the HW3 performance is equivalent
We still need to see if the HW3 performance is equivalent
I really hope you are not trolling..
powertoold
Active Member
Why would I be, first drive was absolutely phenomenal, much much better than V11 would do in terms of comfort, confidence, and human-like decisions / maneuvers
A video is worth a million words in this case, and there were essentially no faults during the first drive (except for one instance of wheel jerk during a pullover at 3mph)
I'm not sure what you are reacting to in this clip.
The car paused while:
- passing a double parked car
- in the marked crosswalk
- whose passenger was actively opening doors and exiting
- with oncoming traffic.
All that just from a single video of less than 5 miles by Omar of all ppl? Ask yourself how many serious safety disengagements from v11 has Omar posted?
How many near miss accidents has Omar ever posted from October 2020? Most people who post as much video as Omar does, have posted maybe hundreds of near accident situations. I can't recall a single one other than the live stream he did with that hyperchange guy. Which isn't surprising, that it occurred in a live-stream because he never posts these disengagements if it were uploaded videos.
He's a complete shill from day one and not surprising he's the first to get public V12 access, because they know he will only post rosy videos.
Last edited:
The only disengagements Omar has posted is when Gali or Chuck was in the car... And then FSDb failed miserably once every 5-10 minutes. I am not trusting him for an honest opinion.
jeewee3000
Active Member
I agree Omar is biased to say the least.
My takeaway is that FSD v12 is one step closer to wide release, which will lead to many drives being uploaded to the interwebs for us to form an opinion over.
Curious to see if actual smart summon is included in this V12 build. (I guess not).
My takeaway is that FSD v12 is one step closer to wide release, which will lead to many drives being uploaded to the interwebs for us to form an opinion over.
Curious to see if actual smart summon is included in this V12 build. (I guess not).
willow_hiller
Well-Known Member
It seems to have gone out to a few unknown drivers as well. I imagine most people aren't set up to record and upload driving video within 12 hours of receiving a software update, but given the demand for them, we might get more videos soon. Until then, only verbal reviews:
Omar says it still is very much beta and not ready for mass customer release. Says it has some regressions and bugs.
I thought it would recognize fire trucks, flashing EMT lights, and add a little caution around that fireman. And then with a firetruck obstructed view it dashed into the intersection like v11.
And I wasn't thrilled with how quickly it crossed over the dashed line and into oncoming traffic when going around double parked cars.
Otherwise, on initial impression, it appears to be smoother but admittedly that's always deceiving in videos.
I thought it would recognize fire trucks, flashing EMT lights, and add a little caution around that fireman. And then with a firetruck obstructed view it dashed into the intersection like v11.
And I wasn't thrilled with how quickly it crossed over the dashed line and into oncoming traffic when going around double parked cars.
Otherwise, on initial impression, it appears to be smoother but admittedly that's always deceiving in videos.
Definitely trolling. Has no real knowledge of v12 performance over previous versions (just like the rest of us). For those of us in reality, we'll just have to wait and see when we get it.
Ok, now i understand. You are referring to Omar's video. I thought that you had it yourself.
Unfortunately Omar doesn't have credibility.
powertoold
Active Member
Yup, that first drive video is all I need to see to know it'll do much much better than V11 on city streets.
It's surprising how seemingly stable and coherent V12 is already.
It's surprising how seemingly stable and coherent V12 is already.
Omar's showed the wheel did not jerked left and right like V11.
diplomat33
Average guy who loves autonomous vehicles
I am not surprised if V12 ends up being better in every day, common driving. End-to-end is a very efficient way of training NN. And NN are very good at imitating patterns. Tesla says that they used millions of clips for V12. So if you plug in millions of clips of good driving, you can except V12 to be good at imitating what it sees in those millions of clips. And I am not surprised if V12's driving is more natural, more human-like. Tesla has trained it on millions of clips, presumably taken from good human drivers, so we would expect V12 to imitate that human behavior and therefore drive in a more natural, human way. We also saw this when Waymo changed to a ML-first planner. They also saw a big improvement in the driving being less robotic, more natural, more confident.
My 2 big questions right now (which I am sure will get answered in time):
1) We know in the past, Tesla was overfitting the data, causing FSD beta to drive better in places like SF and not as well in other parts of the country. Will we see a similar pattern with V12? I look forward to trying V12 in my area where FSD Beta has struggled for a long time.
2) Will end-to-end get to 99.99999% reliability where Tesla can remove the driver supervision requirement and if yes, how quickly will that happen?
My 2 big questions right now (which I am sure will get answered in time):
1) We know in the past, Tesla was overfitting the data, causing FSD beta to drive better in places like SF and not as well in other parts of the country. Will we see a similar pattern with V12? I look forward to trying V12 in my area where FSD Beta has struggled for a long time.
2) Will end-to-end get to 99.99999% reliability where Tesla can remove the driver supervision requirement and if yes, how quickly will that happen?
Similar threads
- Article
