So who's going to be the first to start a whole new thread for 10.69.1.1
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD Beta 10.69
- Thread starter Buckminster
- Start date
Nice, I'm always wondering if 10.12.2 will run into one of those today since they're not rendered and the red line on screen is typically drawn at the curb.
mtndrew1
Active Member
Gosh that’s a lot of the flame. I’ll need to keep an extinguisher in the frunk.So is 10.69.1 going to be(note the extra flame)?
powertoold
Active Member
I'm having trouble believing how far and accurately the occupancy network can see. I measured 400 feet from the ego car to the left-most dining area. The distance prediction is very stable (as you can see in the video). It's beyond anything we've seen with vision-only. Mind blowing. I almost want to think that these predictions are aided by cloud meta-data, but it's unlikely.
The ego car is the red pin, and the orange arrow is the dining area on the left in the video:
The video:
The ego car is the red pin, and the orange arrow is the dining area on the left in the video:
The video:
Even if the FSD was trained to move ahead if the other car stops for you it could be a problem with two FSD cars. Both cars stop -- see the other car is stopped so they both move ahead -- see the other car is moving so both cars stop --- see the other is stopped so they both move ahead. Hmmm.
AlanSubie4Life
Efficiency Obsessed Member
While not exactly the same thing, we know the car in general MUST be able to see at least 400 feet to do what it is doing with unprotected lefts safely (5-7 seconds at 50mph). Stationary objects might be more difficult, but at 150m straight on there is not a lot of movement in a vehicle either.
Anyway it seems reasonable and I would definitely expect it could see this far.
No idea how their occupancy network and other methods of detection are all merged together. I don’t really know much about any of those technical details. All we know is how far it must be able to see at a minimum. Reliably? Who knows…
Hopefully it can see a lot further! Seems like 700-800 feet would be even better, though to some extent resolution limits can be compensated for by faster maneuvers.
Chuck should try his unprotected turns at night sometime when there is a decent amount of traffic! Guess might have to wait until later in the year.
Last edited:
powertoold
Active Member
HW4 would bring higher resolution cameras, so it'll see farther.
The amazing thing about the occupancy network is that it's being deployed to production now. I'm sure Tesla has been working on it for a long time. Elon has been saying V11 is coming for over a year, and it's still not here yet. Tesla has been hard at work, polishing, tweaking, and training these networks to be good enough for public deployment.
I don't fully understand how the occupancy networks are being trained or generated. A lot of it is likely:
1) Using parallax / different focal lengths from the 3 frontal cameras along with radar signals
2) Using IMU / GPS / hindsight data to measure out the distances and volumes of all pixels
3) Etc. (beyond my IQ)
Im currently on 2022.20.8 with my MX which was delivered two months ago when I traded in my M3 with fsd beta. My current safety score is 94.
I’m hopeful that the 10.69 software build number might mean I get fsdb next week, once it’s released widely.
Hopeful but definitely not optimistic knowing how long this process typically takes.
I’m hopeful that the 10.69 software build number might mean I get fsdb next week, once it’s released widely.
Hopeful but definitely not optimistic knowing how long this process typically takes.
Silicon Desert
Well-Known Member
interesting. I haven't gotten it this updated build that is theoretically going out. Then again, I am not exactly an employee
I haven't heard Elon stating that FSD Beta 10.69 will be released to additional beta testers when 10.69.2 goes wide. In fact, He actually said 10.69.2 will go wide to 100,000 people, which is the number of people in FSD Beta right now. Unless we hear otherwise, i would assume you may have to wait a while longer for the next update to additional people.
Are you in the YT celebrity group?interesting. I haven't gotten it this updated build that is theoretically going out. Then again, I am not exactly an employee
I don’t think they are that incompetent. As you also noted, it may be an issue with latency delay. Although, I don’t understand exactly why it appears to take nearly 2 seconds of delay before they apply the brake aggressively. I have noticed a similar braking delay in other contexts such as cross traffic appearing just after my car begins moving into an intersection. That kind of latency may be tolerable in some contexts but for urgent braking due to a pending potential collision that is totally unacceptable. It takes people only a few dozen milliseconds to apply the brakes in similar circumstances. Either the car lacks sufficient cpu or they need to build in a fast path for urgent braking.
Daniel in SD
(supervised)
Well obviously a human would not have to cross double yellow in this particular situation if they were planning on waiting for the car to pass. You stated that it couldn’t see the parked cars behind the parked car which if true is an issue with the camera placement.
j0shm1lls
Member
Keep an eye on the FSD visualization at the timecode in this video - wonder if this has anything to do with the need for a bug fix build or if its more due to a driving characteristic fix.
Last edited:
Looks like the software crashed, ran on backup stack while the FSD stack restarted.
Might be a 10.69.1 revision, might be 10.69.2:
Dogtone
Member
Notice how he reduces the speed? Zero takeovers right?? Classic Omar!
Similar threads
- Article
