Tesla is able to examine the data and say whether the car was on AP or not. I don't even know if other cars can do that. How many deaths on ACC have there been in cars other than Tesla? I suspect a lot but it isn't tracked. AP is ACC with lane keep assist unless you get to NoA. I don't know of any other car where accidents are tracked vs. the use of ACC and lane keep assist. This idea that only Tesla AP has accidents is right up there with only Teslas burn.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
-
Tesla's Supercharger Team was recently laid off. We discuss what this means for the company on today's TMC Podcast streaming live at 1PM PDT. You can watch on X or on YouTube where you can participate in the live chat.
Tesla autopilot HW3
- Thread starter verygreen
- Start date
You are correct here and the only data we have, Tesla's, can be accused of cherry picking.
However, I have been personally impressed. I thought a lane was clear and would have moved into it but since I was on AP I used the turn signal. The car started right, my girlfriend screamed, the car jerked back left, sped up then merged right safely. A car had been coming up that I hadn't seen. I was left very impressed.
I have also seen the limitations. Driving an older AP1 car, I had the car in front move into the right turn lane. The Tesla accelerated even though there was a car stopped in front of me. That car had been uncovered by the car moving into the turn lane. The radar thought the stopped car was part of the road. Same Doppler shift.
Wow this is pretty extraordinary claim There's been no deaths in other cars while ADAS has been engaged.
How has this supposed fact been established?
Every time a Tesla crashes the first question is was AP engaged? My hunch is that if this question was asked of every other car then we would have more data, but to date it's not being asked. I'm going to go ahead and disagree that no deaths have occurred with other L2 ADAS systems. My assumption is they have, we just haven't heard about them. This will change as more and more cars are on the road with L2 capabilities.
Ummm... ADAS level 1 is cruise control... do you really want to claim there have been zero fatalities due to use of cruise control?
The HW3 boards, as I understand it, don't have NVIDIA GPUs. So the new tensor processing chips are already in use. Otherwise, it wouldn't be able to do much of anything. The new neural net isn't in use, but the old NN is running on the new chips.
You'll note that I didn't use the word "randomly" with respect to dropping frames. Most video cameras support only a very limited set of frame rates. Musk says that the current AP2.5 hardware processes 200 fps total. Divide that by 8 cameras, and you get 25 fps. Maybe it is using a PAL-capable camera and running at 25 fps, or maybe it is running the camera at 30 fps and dropping every sixth frame deterministically. The latter would explain a lot of the quirky behavior I've seen, but it is pure speculation; I have no way to know for sure.
Gotcha. If processing starts on the reception of a new frame (as opposed to the most recent staleish one) then it will not skip. Otherwise yeah, there will be a weird beat frequency.You'll note that I didn't use the word "randomly" with respect to dropping frames. Most video cameras support only a very limited set of frame rates. Musk says that the current AP2.5 hardware processes 200 fps total. Divide that by 8 cameras, and you get 25 fps. Maybe it is using a PAL-capable camera and running at 25 fps, or maybe it is running the camera at 30 fps and dropping every sixth frame deterministically. The latter would explain a lot of the quirky behavior I've seen, but it is pure speculation; I have no way to know for sure.
Umm...ya...that's a *not smart* thing to say if you're going to claim your usual position is fact based.
Sorry but this is completely wrong.
SAE Level 1 means the car controls one parameter of the dynamic driving task within the ODD. TACC is Level 1, as is older auto-parking where car handles steering but driver handles accelerating.
Traditional cruise control speed setting is controlled by the human and is therefore SAE Level 0. Just maintaining a speed is not considered automation in the SAE levels, because there is no sustained reaction to the driving environment. (Also some warning or even reactionary type of aids are just Level 0 when there is no sustained control of a parameter like speed or steering adjustment.)
Indeed I’d wager adaptive cruise control or TACC — Level 1 — has been one the most safety-enhancing of all the driver’s aids... I expect much more so than any auto-steer to date.
Last edited:
Humm, one could say it's a crappy level 1 that doesn't recognize moving or stopped objects
. But point taken, let's just go with TACC which is better versus traffic, but just as bad versus black ICE and swerving/ sleepy drivers. Our Explorer's TACC will happily follow another car for miles, then rear end it when the lead car stops (which would likely not be fatal).Remember, the stanard set forth by bladers is zero deaths, not safer.
I guess the Uber fatality was technically outside the car...
That was not what I thought but I could be wrong so I will look into it deeper. Thanks.The HW3 boards, as I understand it, don't have NVIDIA GPUs. So the new tensor processing chips are already in use. Otherwise, it wouldn't be able to do much of anything. The new neural net isn't in use, but the old NN is running on the new chips.
There are probably around 100 million Level 2 ADAS. All of them except super cruise (which uses camera) has a 7-15 secs timer and most don't use the stupid torque sensor that Tesla but touch capacity sensor also the CEOs doesn't hype the product up to be self driving.
This is why there has been 0 death out of trillions of miles driven in ADAS other than tesla.
Now tesla on the other hand started with AP 7.0 or 7.1 which allowed you somethings to go 30 mins to 1hr without placing hands on the wheel. Immediately 2 people died from that. How in the world did you not see that as being dangerous? If other companies did that, you bet ya there would be deaths. Tesla deaths are completely explainable they are simply gross negligence.
No ADAS includes level 2.
There has absolutely been zero deaths because other car manufacturers weren't gross negligent.
Again I repeat, There are probably around 100 million Level 2 ADAS. All of them except super cruise (which uses camera) has a 7-15 secs timer and most don't use the stupid torque sensor that Tesla but touch capacity sensor aswell. Also and more importantly the CEOs doesn't hype the product up to be self driving.
This is in contrast to Tesla who started with AP 7.0 or 7.1 and allowed you sometimes upwards of 30 mins to 1hr without placing hands on the wheel. Immediately 2 people died from that. They then decreased it but not enough and still more people died.
- Do you think if the initial 30 mins were still here would there be more death count or less?
- Do you think if it was 7-15 seconds as other car manufacturers with capacitiy touch that there would be ZERO deaths.
- Do you think if other manufacturers allowed upwards of 30 without placing hands and no way to know you are even paying attention in those 30 minutes, that there won't be deaths?
Remember that all deaths happened because driver fell asleep trusting the system. As Elon said himself its experienced AP drivers that die not inexperienced ones that don't trust the system.
stretchsje
Member
How do we know this? Do we have confirmation that there are a cluster of ARM CPU cores as part of the NN processor? Is the new chip a GPGPU or merely an accelerator of sorts? I've seen nothing verified. It was originally my assumption that the new chip would literally plug into the old board using something like a pci-x slot, thus making the upgrade path extremely simple and thus field-deployable, but I think there are a multitude of reasons why that assumption was probably wrong.
The new chip is a tensor processing unit that is specifically designed to run machine learning models. It serves the same purpose as the GPU in the older hardware, but is faster at it because it is tailored to that purpose. It would be redundant and expensive to add NVIDIA chips on top of that, and the move from an NVIDIA SoC to a Samsung SoC makes it even more unlikely that they would do so.
boonedocks
MS LR Blk/Blk 19”
Even if they could / do I am not certain it would make the news and Twitter and every socially available media. When you are in the forefront you are also in the microscope. Accidents with legacy automobile manufacturers cars are...well....just accepted as daily occurrences.
Source? Got board pics you can PM me? The new neural net isn't in use, I call BS, it is on dev cars somewhere...
And if some kind soul that wants 5 mins of fame and has a torx screwdriver that got an APH4 delivery will PM me, we can stop speculating
Well not really, Level 1-2 have the driver in charge, so there are many things they don’t recognize. Even a Level 2 system that recognizes stopped cars will not always recognize debris etc.
That is why the ODD (operational design domain) is defined for each feature on at least SAE Levels 1-4. It can be a speed range, it can be geographical area, it can include limitations etc.
Last edited:
Source? Got board pics you can PM me? The new neural net isn't in use, I call BS, it is on dev cars somewhere...
And if some kind soul that wants 5 mins of fame and has a torx screwdriver that got an APH4 delivery will PM me, we can stop speculating
It would make sense that the HW3 boards not have Nvidia GPUs, only tensor processors and CPUs, unless there is some transitional plan of limited time.
Here's my take on accidents in Teslas. If (E)AP is on shame on the driver. None of that stuff is GA (General Availability). It's use 'requires' active driver participation. At least in the Model 3s, without an indication the driver is paying attention the car will drop out of (E)AP. I don't know if this is true for S and X.
Then, at the risk of sounding what we in the UK call a bell end, it is not Tesla's fault. Its the fault of the driver/s that went to sleep.
People fall aslepp at the wheel in all sorts of cars. They die asleep at the wheel in all sorts of cars. The fact that a Tesla makes a system which makes driving a less intense activity and therefore gave them, and you it would seem, a poor excuse to fall asleep, does not exhonerate them from blame.
If you feel sleepy behind the wheel you pull over, whether you are driving a Tesla or a tractor.
Similar threads
- Article
