-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
I don't know, that just seems very optimistic. That in 3 months, it'll go from the crap it is now to significantly improved? We've been on FSD beta for almost 2 years now, and it's nowhere near good. Every update is 2 steps forward, one step back, sometimes even 2 steps back, one step forward lol. They still can't figure out how to slow down without friction brakes for a red light, they still put turn signals on for sharp curves, tons of lane assignment issues as well. If they still can't get the basics right, how will they be 90% better than a human in 1 more year?
Daniel in SD
(supervised)
Nope, L3. No reporting is required for L2.
12.2 miles of L3 testing in California in 2019. They do also claim to do testing in jurisdictions that don’t require reporting.
I think what we understand as fully released will still be L2 assist but still classified as a beta.
NOA is actually a beta feature that you have to accept the terms of to enable. A lot of people forget that.
I think that it will just be something you can buy and enable and will always be a "experimental WIP" with no tesla responsibility of fault. BUT.. that's not to sat it won't be incredible in a few years. I just think that it will never be a feature like heated seats, it's always going to have terms and conditions that tell you to be aware at all times even though it may not nag you anymore
Maybe it eventually becomes L3 or L4 in time, but it will first release to the public as a opt in "open beta like NOA
NOA is actually a beta feature that you have to accept the terms of to enable. A lot of people forget that.
I think that it will just be something you can buy and enable and will always be a "experimental WIP" with no tesla responsibility of fault. BUT.. that's not to sat it won't be incredible in a few years. I just think that it will never be a feature like heated seats, it's always going to have terms and conditions that tell you to be aware at all times even though it may not nag you anymore
Maybe it eventually becomes L3 or L4 in time, but it will first release to the public as a opt in "open beta like NOA
DarkForest
10.69 made me a shill
There’s no such thing as partial/occasional L5. L5 is 100%. Any sort of segmentation can be at best L4.
There’s also no difference in terms of L3, L4, and L5 liability. If the car is driving itself and hasn’t told the driver they must take control within X number of seconds (exception not applicable to L5 obv), then Tesla is liable, period.
Liability is the whole point of L3 and beyond, and I don’t ever see Tesla taking on that liability in order to sell a feature, especially not one that’s being widely purchased as it is.
Another way to look at it is:
L2: hands off
L3: eyes off
L4: mind off
L5: steering wheel optional
I personally think the current hardware will never go past L2. I’d bet a dozen tacos that at best, we’ll get interstate highway L3 as long as you have Tesla insurance. I’d bet two dozen tacos that Musk will eventually try to pass FSD off as “essentially L5 if it wasn’t for those pesky regulations [that require Tesla to be liable for collisions which is whole point of L3+].”
Attachments
diplomat33
Average guy who loves autonomous vehicles
Everything in your post was really good. This is the only part I would take issue with. I don't really like this way to describe the levels. It attempts to describe the levels based on the amount of attention the human needs to give but it is not very accurate IMO. Here's why:
L2 = performs both steering and braking tasks but requires a human driver to perform any missing driving tasks.
Some types of L2 like SuperCruise are hands-free in some situations because they rely on camera based driver monitoring. But not all L2 are hands-free. And L2 is certainly not hands off by default as it requires active driver supervision and intervention.
L3 = performs all driving tasks in a limited ODD but human driver is the fall-back.
So yes, in some instances, the human can take their eyes off of the road but usually for a limited amount of time. L3 can require the driver to take over. L3 is not eyes off all the time.
L4 = performs all driving tasks in a limited ODD with no human fallback.
Yes, human can take their mind off of driving. But the steering wheel is also optional for L4 too, ie Waymo firefly, Cruise Origin, Zoox robotaxi. So the steering wheel being optional is not just a L5 thing.
L5 = performs all driving tasks in the entire "human driver ODD" with no human fallback.
Yes, the steering wheel is optional but that is not the key difference between L4 and L5. L5 is also mind off. The only difference between L4 and L5 is where and when they can operate.
Last edited:
DarkForest
10.69 made me a shill
Yeah, that’s legit. I found that from a Wiki page and thought it was close enough.
The way I prefer to think about it is 1. liability and 2. how long do you have to take control when the car asks you to. However, I didn’t find a good resource with this info to post.
Based on memory (probably incomplete/inaccurate):
L2: driver must always pay attention and be ready to take control at all times, even if the car doesn’t ask them to (eg it’s about to do something wrong). Driver is always liable
L3: driver does not need to pay attention unless the car asks them to, then the driver will have sufficient time to take control and understand their environment (eg 15 seconds). Tesla would be liable while it drives and in those 15 seconds
L4: driver does not need to pay attention unless the car leaves the geofenced area (location-based rather than time-based, so it could be “you will need to take control in 25 minutes when we leave the area.”). I think there could also be other requirements for going into L4 like weather and time of day; I’m not sure. Tesla would be liable in the geofenced area and when the car is coming to a stop at the edge of the area if the driver does not take control
L5: car can handle all driving situations, absolutely zero exceptions. Tesla would be liable except when the car is being driven manually assuming it even has that option
diplomat33
Average guy who loves autonomous vehicles
Yeah that is pretty good. You are correct about liability.
A couple observations:
Yes, L4 can be limited by other factors like weather, speed, time of day. For example, we see Cruise AVs are limited to only operating at night. ODD involves many factors including types of roads, weather, time of day, speed, geography, traffic condition.
L5 is not expected to handle all driving with zero exceptions. SAE says L5 is only expected to drive in conditions that a skilled human can be expected to drive in. So there are weather conditions like a white out snow storm that L5 would not be expected to drive in since humans are not expected to drive in it either. L5 is expected to automatically pull over when conditions are not drivable and wait for conditions to improve.
Here is the full paragraph from the SAE J3016 (section 5.6, page 32, note 1):
"“Unconditional/not ODD-specific” means that the ADS can operate the vehicle on-road anywhere within its region of the world and under all road conditions in which a conventional vehicle can be reasonably operated by a typically skilled human driver. This means, for example, that there are no design-based weather, time-of-day, or geographical restrictions on where and when the ADS can operate the vehicle. However, there may be conditions not manageable by a driver in which the ADS would also be unable to complete a given trip (e.g., white-out snow storm, flooded roads, glare ice, etc.) until or unless the adverse conditions clear. At the onset of such unmanageable conditions the ADS would perform the DDT fallback to achieve a minimal risk condition (e.g., by pulling over to the side of the road and waiting for the conditions to change). "
The SAE also allows for business or legal limitations to L5. So L5 can have ODD limits as long as they are based on business or legal reasons. The limits just cannot be technological.
Here is the full paragraph on that (section 8.8, page 39):
"There are technical and practical considerations that mitigate the literal meaning of the stipulation that a Level 5 ADS must be capable of ‘operating the vehicle on-road anywhere that a typically skilled human driver can reasonably operate a conventional vehicle,’ which might otherwise be impossible to achieve. For example, an ADS-equipped vehicle that is capable of operating a vehicle on all roads throughout the US, but, for legal or business reasons, cannot operate the vehicle across the borders in Canada or Mexico can still be considered Level 5, even if geo-fenced to operate only within the U.S. The rationale for this exception is that such a geo-fenced limitation (i.e., U.S., only) would not be due to limitations on the technological capability of the ADS, but rather to legal or business constraints, such as legal restrictions in Canada and Mexico/Central America that prohibit Level 5 deployment, or the inability to make a business case for expansion to those markets."
So the bottom line is that L5 has to drive everywhere that a human can and within any business or legal limits. So for example, you could have L5 that is US only and it would need to be able to drive everywhere in the US that a human could be expected to drive.
By the way, 2 years ago, I created a "primer" on the SAE levels where I attempted to summarize the key points in the SAE J3016 standard. The thread has lots of good information that you might find useful:
Primer on SAE Levels of Autonomy
Since I know we have a lot of discussions about the SAE levels, I thought I would write up a primer from the SAE document: General The levels are not in order of quality, or merit or technology. L5 is not necessarily “better” than L4. L4 is not necessarily “better” than L3. A “high...
teslamotorsclub.com
Last edited:
DanCar
Active Member
All answers are possible, just go further into the future for more functionality and set definition of what fully released means. I'll vote for level 4, in 10 years. Level 5 in 20 years. And yeah, it won't be with current hardware.
Last edited:
Tam
Well-Known Member
That's the thing. Tesla said it had all hardware it needed for FSD in 2016 but hardware is a moving target: MCU, Cameras, Computer hardware version 2.0, 2.5 ,3.0 and soon 4.0! Radar is off but who knows Radar will be on again. LIDAR is off but who knows if Elon will no longer be CEO, it might be on at that time...
Last edited:
My WAG.... Unless something dramatically changes with the system design, HW, and software development cycle, FSDb will progress too slow, be too crude, and lag-like-a-distracted driver for any significant gains on the vehicle automation scale. I wouldn't be surprised if Musk is content with lowered long term FSD expectations (L2) for the Tesla lineup and be forced to emphasize an arguably more achievable geofenced robotaxi path. But what do I know.
Olle
Active Member
I voted L5, but didn’t say when.
At one point is time L5 will be a commodity. FSD beta will reach that level between now and that point.
To put it in perspective; If the Singularity happens in 2045 as Kurzweil predicts, L5 will happen much sooner and be beyond obsolete by then.
At one point is time L5 will be a commodity. FSD beta will reach that level between now and that point.
To put it in perspective; If the Singularity happens in 2045 as Kurzweil predicts, L5 will happen much sooner and be beyond obsolete by then.
Daniel in SD
(supervised)
If the singularity occurs why will AI want to drive us around?
My hypothesis is that Tesla already has AGI but just can't figure out how to make it agree to drive. That's why Elon is was so confident at Autonomy day that L5 robotaxis would be ready by 2020. It also explains why he's so worried about AI safety, the AGI they have is very angry.
DarkForest
10.69 made me a shill
“We are excited to announce that, as of today, all Tesla vehicles produced in our factory – including Model 3 – will have the hardware needed for full self-driving capability at a safety level substantially greater than that of a human driver.”
Full Self-Driving Computer Installations | Tesla Support
While no Tesla vehicles are fully autonomous today and require active driver supervision, the FSD computer is capable of delivering intelligent performance and control to enable a new level of safety and autonomy, without impacting cost or range. Learn more about FSD computer installation.
“Will the FSD computer make my car fully autonomous?
Not yet. All Tesla cars require active driver supervision and are not autonomous. With the FSD computer, we expect to achieve a new level of autonomy as we gain billions of miles of experience using our features. The activation and use of these features are dependent on achieving reliability far in excess of human drivers, as well as regulatory approval–“
They were careful to say FSD capability and FSD computer. They also set the standard to just being safer than a human.
I still think they might try to wiggle out of everything by saying FSD is a product name, not a product description, and that’s why they’ll introduce new hardware tomorrow which won’t be retrofitted onto the current fleet. Of course, they won’t immediately say they won’t do retrofits. It’ll get mentioned after the new hardware is available for sale. That way it doesn’t tank current sales.
diplomat33
Average guy who loves autonomous vehicles
You are joking, right?
Nobody has AGI yet. Certainly, if Tesla had AGI, they would not hide it. Elon would be the first to talk about it considering how he loves to boast about "solving difficult AI problems". And AGI would not get angry or refuse to drive. AGI would not have a mind of its own. AGI just means an AI that can do very different tasks. For example, if the same AI could draw a picture, write a piece of music, assemble a puzzle, and solve a math problem, that would be AGI. We are not close to AGI. The most likely hypothesis for why Tesla has not released L5 robotaxis, is simply that Elon completely misjudged how hard FSD is and Tesla has not solved L5 yet.
legendsk
Member
I will not agree that is "fully released" until it meets the stated specifications when I paid for it. IMHO it cannot be called fully released unless it can be issued a driver's license?
Elon may be able to pay for terra forming Mars and building 1000's of rocket ships to fly people there out of petty cash, but there isn't enough money in the universe to pay the bribes required to get the governments of the world to agree that cars are intelligent, have the same rights as people and issue them a driver's license.
Elon may be able to pay for terra forming Mars and building 1000's of rocket ships to fly people there out of petty cash, but there isn't enough money in the universe to pay the bribes required to get the governments of the world to agree that cars are intelligent, have the same rights as people and issue them a driver's license.
DanCar
Active Member
Don't worry , I'm developing AGI as we speak. I'm working on the 12 point alberta plan. I'm modeling it as myself, to be a soldier / slave of the people.
Daniel in SD
(supervised)
I'm just saying it would fit the known facts. Elon isn't letting the world know because the AGI they created is dangerous!
Yes, I'm joking (I think).
It seems like consciousness could emerge from a machine with human intelligence, especially if we're talking about the singularity where the intelligence would be self-improving.
They issue licenses to robotaxis already without giving them the same rights as people. haha.
Similar threads
