3 months maybe, 6 months definitely
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Three months maybe, six months definitely...
You beat me by 30seconds! We'll done!!
With regards to disengagement reporting requirements, the most concrete piece of evidence we have is that Tesla DID report disengagements for their FSD video attempts last year. So to me, without new, stronger evidence, I find it most likely that Tesla did not do similar testing on California roads this past year (ending Nov 30).
With regards to FSD progress, it seems very likely that highly detailed mapping is necessary for Tesla's FSD. Without these maps, road testing is useless. I don't know how Tesla is building (or buying) the maps, but given that @verygreen didn't see them in his car before December, it seems likely to me that Tesla was not doing much (or any) public roads testing of their FSD code (shadow mode or otherwise) before December 2017. Although it is certainly also possible that they have some special development cars that do have HD maps in some places outside California.
Personally, I do believe the reports that Tesla have suddenly commenced testing for the cross-country FSD video, shortly after uploading HD maps in late December. At this time, near parity of AP2 may have freed up some people, and I would also not be surprised if they intentionally delayed start of testing a bit to avoid 2017 California reporting requirements.
With regards to FSD progress, it seems very likely that highly detailed mapping is necessary for Tesla's FSD. Without these maps, road testing is useless. I don't know how Tesla is building (or buying) the maps, but given that @verygreen didn't see them in his car before December, it seems likely to me that Tesla was not doing much (or any) public roads testing of their FSD code (shadow mode or otherwise) before December 2017. Although it is certainly also possible that they have some special development cars that do have HD maps in some places outside California.
Personally, I do believe the reports that Tesla have suddenly commenced testing for the cross-country FSD video, shortly after uploading HD maps in late December. At this time, near parity of AP2 may have freed up some people, and I would also not be surprised if they intentionally delayed start of testing a bit to avoid 2017 California reporting requirements.
While my car did not get the offer to get the maps, the code to fetch them was in place for quite a while so I traced it and got a glimpse. They do have maps, just only for California and very spotty at that (and a separate coast-to-coast repo that started to be filled by end of December).
As such with maps in place it's kind of unlikely that they got the maps, but did not use them. And while some of the maps are for private area (e.g. the truck-specific maps), some were for quite public places too.
Any current car with recent firmware can download maps, you just need to enable the capability by a special switch that is not present in the car, but could be controlled by the mothership (dev cars also have the switch it looks like, but I only have very cursory understanding of how it works since it's not 100% present in production firmware)
hacer
Active Member
It's worth pointing out that Zoox is and has been a major player in the self driving world. They are making their own autonomous cars that have a lot of out-of-the-box thinking (symmetrical front/back (i.e. no special "forward" direction), external airbags that fire on predicted collision, and more) for ride-hailing transportation. But they intentionally remain out of the limelight.
https://spectrum.ieee.org/transport...he-robotaxi-startup-taking-on-google-and-uber
I think it is important to consider that of the 2015,2016, and 2017 reports, there is only one month (and only 5 days in that month) of reported data. That is why I draw a potential distinction between FSD + nag and real FSD with full hands/ feet off operation. For recording an FSD video, they need to be in hands off mode. For the rest of the time, (as far as I can deduce) they can test any software they want as long as it requires a driver torque input (for their purposes, this does not matter as long as the driver does not apply enough force to cause AP to disable).
Thank you, that is extremely useful and interesting. So they've had spotty California HD maps for some (undetermined) time, and they could have potentially used this prior to December 2017 for testing that they did not disclose...
Good point. So perhaps we can only conclude with confidence that Tesla had not started filming the cross-country FSD trip before Dec 2017. Honestly, I'm not sure exactly why Tesla's original autopilot (which did not require hands on wheel) did not count as self-driving testing but Uber's testing did. But I'm also not convinced that driver torque input alone can make such a big difference for regulatory requirements. If so, it seems to me like it would be too easy for Waymo or Uber or anyone to simply require a hand resting on the steering wheel and avoid regulations. Then again, self driving is a new and rapidly evolving technological and regulatory environment, so maybe I shouldn't expect to deduce all the right answers based on a few principles that seem logical to me.
That is also my conclusion.
Opinion:
Waymo and Uber need fully self driving for their product. Waymo is doing full autonomous testing and reporting. Uber tried doing full autonomous (hands off) testing but claimed it wasn't. Tesla is doing similar software, but because they sell cars to people, and not just cars as an auto-taxis, they don't have the need to show hands off driving now. Tesla's sales/ revenue stream is covered anyway. So no advantage to Tesla to try full hands off mode until the cameras are rolling.
stopcrazypp
Well-Known Member
Note that Tesla never claimed AP1 would be self driving capable or fully autonomous. Even when it did not have the hands-on-wheel nag, Tesla always said the driver is responsible for monitoring, and the system was firmly developed to be a level 2 vehicle.
Uber had a joint press release with Volvo about developing autonomous cars (not just "semi-autonomous") before they started testing them. When California pushed them for a permit, then they used the requirement of engineers to monitor the system during testing as a excuse (that it's level 2 like Tesla), but any person can easily see they were developing the cars to be self-driving, not just level 2 (the goal of Uber was always to eliminate the driver).
Volvo Cars and Uber join forces to develop autonomous driving cars
The other thing is Tesla got a license, while Uber didn't. It's blatantly obvious to the DMV when you don't have a license vs just failing to report disengagements (and as I mentioned you can tell from blocks away that the Uber Volvos are self-driving, so eyewitnesses will easily report them).
Last edited:
Well, if it were BMW, for example I'd get it, but CA is where all of the AV engineers are.
And reporting has benefits, if you are good at it:
1) higher stock price, easier loans etc.
2) more people buy EAP and FSD, which is free cash for Tesla, since the HW gets built in anyways
3) the general perception as a tech leading company can boost all kinds of other sales, no matter if they are related to AP, like power walls and solar cells.
Reporting a bad result is always bad, I agree on that. But 0 miles also doesn't sound too good and only real Tesla friendly media spins this in a good way ("shadow mode", simulations)
pdxgibby
Software Engineer
Tesla doesn’t have a problem with demand or stock price or market value. They also don’t have a problem with FSD option. People will buy it when it’s ready. They are sitting on billions in upgrades once it’s ready. It’s better for them to NOT sell it at delivery, as it’s more expensive afterwards.
Reporting nothing is ... stealthy?
Tesla is a company that is loosing money quickly, while having ambitious targets on future growth, so money they have right now is more worth than money in the future. The same goes for the stock price. So it is better to sell now, that why it's CHEAPER for you to buy right now.
And sure Tesla has no problem with Model 3 demand, I'm not sure if they wouldn't like the extra S and X produced.
Investing money in infrastructure is not the same as losing money. I'm also fairly sure S/X are production limited.
Adaptive cruise control requires EAP, so that has a large take rate (portion flammable as EAP is still up for debate). I don't expect FSD sales to increase much until 1. It is released or 2. It transfers with owner. If people could take it with them to next car/ lease it would be more worth doing vs trying to get 33% or so return on the price different by investing.
did you miss the leaked internal documents from Tesla where it became obvious FSD uptake between buyers was huge? I don't remember the uptake right now and cannot find the article on electrek about it either, but it was quite a bit higher than I expected.
Edit: found it - Over 35,000 Tesla owners bought ‘fully self-driving’ feature despite still being unavailable, sources say
40%! (the source is legit)
Similar threads
