Well, I don’t know if it’s dangerous. I follow the rules, both hands on the wheel and ready to take over. If one does that and remembers that you are always the driver I don’t think it’s dangerous.
Jmho.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD discussion
- Thread starter Buckminster
- Start date
Jmho.
Maybe you could make this your signature or something, so you don't have to type it out every....single......week.
Think you are thinking of someone else. Or just exaggerating.
yup ... you can also add the picture of your camper trailer as your avatar to go along with the constant FSDb bashing ...tiresome ...
you really should have your vehicle checked .... your experience is so far from what i have experienced with FSDb .... you must have a different product
then you wont need to post
Last edited:
Lol. Maybe. But pretty much everybody we know has the same results with FSD. Glad yours is working at that autonomous level though.yup ... you can also add the picture of your camper trailer as your avatar to go along with the constant FSDb bashing ...tiresome ...
you really should have your vehicle checked .... your experience is so far from what i have experienced with FSDb .... you must have a different product
then you wont need to post
Close enough that Tesla won't need to design a Gen 4 vehicle platform for Robotaxis, the upcoming Gen 3 will likely still be going strong into production once autonomy is solved (a few years from now).
willow_hiller
Well-Known Member
Lol. Maybe. But pretty much everybody we know has the same results with FSD. Glad yours is working at that autonomous level though.
Model S/X, by any chance? Given how much Tesla has had to retrain for HW4, I'm guessing their models are very sensitive to camera positioning. So my working theory is that they have had to train separate iterations of their models for S, 3, X, and Y; and less training data for S/X has led to worse performance compared to 3/Y. This would also explain why Tesla has tended to equip their lower volume vehicles with more sensors. The HD radar for HW4 S/X may be there to make up for this data volume deficiency.
Knightshade
Well-Known Member
It was 6 years from when Tesla first said they were talking to OEMs about giving them access to superchargers (June 2017) until those first NACS deals were announced in June of 2023.
And it's now 7 years later and that dream still isn't here. Or, as most of us who are testing fsdb can tell you, especially close. Especially with 99.9% of the existing fleet being pre HW4.
As Elon once said- "We specialize in making the impossible merely late"
Last edited:
Sure... I'm just not sure there's a TON of folks in that bucket? Especially when they'd be getting the pre-refresh one.... so the tradeoffs (other than getting a fresh warranty) are somewhat hit and miss.... (heat pump, but no USS or default homelink or adjustable lumbar or center-console data ports for example... and anyone with things like accel boost or PPF are out thousands of bucks to get those on the new one as well)
It is worth something because Tesla promised me my car as purchased was enough to do it.
Then when it turns out it wasn't they gave me a free HW3 upgrade.
if it's still not enough they pretty clearly still owe me what they promised.
(buyers after March 2019 are not in this position, as they were promised far less... but those before then are)
If they utterly fail to deliver what's promised the check is gonna be a lot bigger than $50. At bare minimum it'd be what I paid for FSD plus interest- and there's a pretty solid chance it'd be what FSD cost at settlement on a new (actually has working FSD) car... or have an alternative of exactly this type of new-car transfers but with a lot better terms too.
I mean even the EAP people, who got EVERYTHING they were promised in their purchase, but it was a few months late, got more than $50
Anyway to avoid mod ire, this topic might be better here?
Can you remind me what was promised for hw 2.5 customers when we bought FSD before the language change? Was L3 and beyond promised or it was always L2 FSD with aspiration to be L5? Do you remember Tesla ever promising to take liability for FSD which is the only difference between current L2 fsd and any level beyond.
Knightshade
Well-Known Member
HW2.0 and HW2.5 customers were absolutely promised a feature set that is at least L4 by SAE definitions (because by definition that it is the lowest level the car can operate without any action required from a person who is sitting in the drivers seat- and Tesla promised the car would be able to complete short and long trips in exactly that way)
There's an argument the promise is for L5-- but I think in a court, with good lawyers, L4 with a "reasonable" large ODD would satisfy the promise.
The only disclaiming of the promise was it was dependent on regulators (which is already a non-issue as it's legal TODAY in a bunch of US states if the system worked) and on "extensive software validation"-- which is "We need to be sure it works" not "We hope to someday invent something that does work aspirationally"
You can see the promise in question here:
(and just to head off/save you the already repeatedly debunked arguments from the FSD forums- "action" both in the dictionary and the SAE standard includes monitoring the vehicle and it includes remaining ready to, and of course actually, taking over driving it- so yes, this is a minimum of L4 promised here)
If was only ~April 2019 when Tesla completely revamped the description of FSD and it's promises and features to only promise an L2 system, with some aspirational language around it. They also moved all the EAP features into FSD at this time.
FSD as sold pre/post 4-19 is a fundamentally different product and a fundamentally different promise made to the buyers of each product.
There was considerable discussion in here- I'm surprised you don't recall- that the change was explicitly to reduce Teslas liability in a future court case if they fail to deliver >L2.
This is 100% false and has ALSO been debunked repeatedly in the FSD forum.
Nothing in the SAE levels even uses the word liability, and does not care about who, legally, takes liability for anything.
It cares about the capability and design intent of the actual system. Can the system perform the ENTIRE DDT without EVER requiring action from a human?
If yes, and it has no ODD, it's L5.
If yes, and it DOES have an ODD, and it can fail safely without a human if it leaves the ODD or has a system failure, it's L4.
*** This is the minimum Tesla promised above to 3/19 and earlier FSD buyers
If yes, and it DOES have an ODD, and it can NOT fail safely without a human if it leaves the ODD or has a system failure, it's L3.
***This is what Mercedes is currently delivering to customers in parts of the EU and a few US states today
If no, it's at best L2.
***This is where we are today with FSD- because Teslas system has an incomplete OEDR (per Tesla themselves in legal filings with CA DMV) so it can not perform the entire DDT without human action.
Liability don't enter into SAE levels. At all.
Once again, there's a whole forum, with lots of threads that already have deep dives on all of the above, for further discussion:
AI, Autopilot, & Autonomous/FSD
Discussion about AI, Tesla Bot, Tesla Autopilot (AP), the promise of Full Self Driving (FSD), as well as other Autonomous Vehicles.
teslamotorsclub.com
"If yes, and it DOES have an ODD, and it can NOT fail safely without a human if it leaves the ODD or has a system failure, it's L4."
Why is this a contradiction statement from
"Can the system perform the ENTIRE DDT without EVER requiring action from a human?"
If it can get to a destination without ever requiring a human action, why would it ever fail?
We see many waymo and cruise failed to get to their destination, are they no longer l4?
Why is this a contradiction statement from
"Can the system perform the ENTIRE DDT without EVER requiring action from a human?"
If it can get to a destination without ever requiring a human action, why would it ever fail?
We see many waymo and cruise failed to get to their destination, are they no longer l4?
Knightshade
Well-Known Member
The line you quote has L3 at the end now- I did edit it a couple times right after posting you might've caught something before that where L4 got repeated.
If you look now it's correct- my apologies for any brief confusion.
If it has an ODD it's always possible something will happen outside its ODD during the trip.
The difference between L3 and L4 is if the system can handle that possibility WITHOUT a human.
That depends on how they handle the failure. They are designed as L4-- that is they "fail safely"- what SAE calls minimal risk condition.
In contrast Mercedes system is L3-- if it fails it REQUIRES a human to be there to insure safe operation- it can not handle a failure on its own safely.
Again that's what 3 vs 4 is primarily about.... and why you can sleep in an L4 car (or not even physically have any humans in it) but in an L3 car you must have a human present in the drivers seat, AND they must be actively awake, and prepared to take over the driving task on relatively short (but not immediate) notice.
I see many times when physical people from cruise and waymo take over their cars and drive away due to unforseen problems. And if a human was in the cruise then it wouldn't have hit the bus so there goes that "no human is required for failiures" definitionsl. So I feel like all of these definition are asinine and like the liability qualification much better.
Last edited:
Knightshade
Well-Known Member
You mean after they failed safely without a human and pulled over? More often this is handled remotely now, though still not always.
The fact they sometimes need a human to do something with them after that doesn't change anything though. If they were not L4 they couldn't ever operate safely without someone in the drivers seat.
That's not the actual words I used of course. I spoke to when a human is required to perform some aspect of the systems functions. The "DDT fallback" role is an aspect of the system. In L3 DDT fallback is by design a human so a human is required to be there and able to perform that function. In L4 by design the system is meant to perform the DDT fallback WITHOUT requiring a human (though one can still be there, it just can't require them to be).
That doesn't mean the system always works perfectly 100% of the time- no system does. It means it's capable of doing so, and designed to do so as fully as possible.
You don't seem to have actually read J3016. You really should before getting further into the weeds here.
You can "feel" whatever you like, but there's a few fact-based things to be aware of:
Your actual understanding of the SAE levels appears to be.... not especially strong? I HIGHLY suggest you take the time to read J3016 since you're still getting some basic stuff wrong. You can grab a free copy here:
J3016_202104: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles - SAE International
This document describes [motor] vehicle driving automation systems that perform part or all of the dynamic driving task (DDT) on a sustained basis. It provides a taxonomy with detailed definitions for six levels of driving automation, ranging from no driving automation (Level 0) to full driving auto
Your feelings are irrelevant to the law. Legal liability is NOT an aspect of the SAE levels.
The actual law, everywhere there is one currently, uses the SAE levels and definitions- so that's where things are today.
The actual functionality Tesla promised pre 3/19 buyers is, at minimum, L4. They currently are at L2.
Last edited:
You posted what Tesla promised. I don't remember seeing the phrase "system failing does not require human intervention". It just says "it can take you to your destination without driver intervention". Did they promise no human fail back by design in what they promised?You mean after they failed safely without a human and pulled over? More often this is handled remotely now, though still not always.
The fact they sometimes need a human to do something with them after that doesn't change anything though. If they were not L4 they couldn't ever operate safely without someone in the drivers seat.
That's not the actual words I used of course. I spoke to when a human is required to perform some aspect of the systems functions. The "DDT fallback" role is an aspect of the system. In L3 DDT fallback is by design a human so a human is required to be there and able to perform that function. In L4 by design the system is meant to perform the DDT fallback WITHOUT requiring a human (though one can still be there, it just can't require them to be).
That doesn't mean the system always works perfectly 100% of the time- no system does. It means it's capable of doing so, and designed to do so as fully as possible.
You don't seem to have actually read J3016. You really should before getting further into the weeds here.
You can "feel" whatever you like, but there's a few fact-based things to be aware of:
Your actual understanding of the SAE levels appears to be.... not especially strong? I HIGHLY suggest you take the time to read J3016 since you're still getting some basic stuff wrong. You can grab a free copy here:
J3016_202104: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles - SAE International
This document describes [motor] vehicle driving automation systems that perform part or all of the dynamic driving task (DDT) on a sustained basis. It provides a taxonomy with detailed definitions for six levels of driving automation, ranging from no driving automation (Level 0) to full driving autowww.sae.org
Your feelings are irrelevant to the law. Legal liability is NOT an aspect of the SAE levels.
The actual law, everywhere there is one currently, uses the SAE levels and definitions- so that's where things are today.
The actual functionality Tesla promised pre 3/19 buyers is, at minimum, L4. They currently are at L2.
What am I missing here?
Knightshade
Well-Known Member
You keep quoting things neither Tesla NOR the SAE said.
No, it definitely DOES NOT say that.
Why do you keep misquoting?
andWhat Tesla actually said said:
What Tesla actually said said:
That's (at least) L4, by definition- most relevant bits in bold. (You can argue for L5, but that's much murkier an argument- L4 minimum is very clear here because below L4 you always are required to take some actions beyond getting in and inputting a destination).
I covered that in the original post.
They state "no action required" from the human in the seat. Not "no intervention" which is a very specific action- but no action period
They state "all you will need to do is get in and tell your car where to go"
All levels below 4 by definition requires action from the driver and require you doing more than JUST getting in and telling it where to go.
Specifically L3 requires you must remain alert (an action) and be prepared (an action) to (when asked) perform the DDT fallback task (also an action). That third one you may or may not have to actually take during any given drive--- the first two you must always take on every drive.
This is why you can go to sleep in an L4 car and can not in any level below that.
L2 requires all those things AND for you to perform ADDITIONAL actions to fill in whatever additional elements of the DDT the system is not capable of doing by itself (in the case of FSDb for example, per Teslas own words, the system can not perform the complete OEDR task- so the human must take action and perform that task- in addition to performing all the ones they would in an L3 system)[/QUOTE]
Last edited:
JRP3
Hyperactive Member
Telsa may someday solve anti-gravity but I don't want to hear Elon talking about it until it works as intended, and neither does Wallstreet.
Well you can't blame Elon for not being transparent. He wants to move volume at any margin so FSD can make it up afterwards. Not a business plan wallstreet agrees with. If FSD works someday, what Elon is doing would be the right business move. If it doesn't then well...I guess 10% sell off tells me wallstreet is betting that it most likely doesn't.
Similar threads
