Elon: "Feature complete for full self driving this year"

[Eno Deb]

As for what SAE level it is...who the F cares. If I can get into my car, tell it to go to Stop & Shop (a local grocery chain) and worst case all I have to do is put about 1 lb of torque on the steering wheel to signify I am paying attn, I'll call that FSD. If I travel from New England to NJ or MD or PA and it successfully navigates the George Washington Bridge Plaza I'll call that self driving too.

What's the point of "full self-driving" if I can't use the time to do some work or watch a movie because I have to supervise the car? I'd also like my Robocar to be able to drive without me, e.g. to come pick me up at the airport or park itself somewhere.

 

[Daniel in SD]

Huh! That's surprising. To lose nags and driver attn (lawyers, if no one else, would demand for legal reasons) would require regulator approval or Congress. As I said, for my purposes, if the car will get me from point A to point B without me being the physical driver, I am OK with attentiveness and I'd consider it FSD or close enough for government work. Consider this though. You and I might disagree on FSD or why but, IMHO, the real big drivers of what'll happen are UPS, FedEx and the long haul companies. There is money to be made by eliminating the trucker, limited by regulations to only so many hours / day...breaks, sleep etc. If they can encourage Tesla for a solution (their tractor trailer cab) they'll pay big. So, I think it's game-on. I think a lot will happen in the next 6 months. I could be wrong but there is serious money on the table now.

Personally I'd prefer to have a car that allows me watch a movie or read a book...
There has been serious money on the table for quite a while now (I've read $80 billion+). I wouldn't hold my breath, it's an extremely difficult problem.

 

[wcorey]

What's the point of "full self-driving" if I can't use the time to do some work or watch a movie because I have to supervise the car? I'd also like my Robocar to be able to drive without me, e.g. to come pick me up at the airport or park itself somewhere.

That'll come. I am merely saying if the car does what I described I'll call it FSD regardless of some engineering book definition. I believe the relaxation of nag and please pay attention is more legal or, will become more legal than technical. The guy reading a book or sleeping in back seat who's S ran into a lane divider would have won a multi-million dollar suit had it not been for the beta designation and 'you must pay attention and be able to take control at any time". Ditto several other tesla fatalities. When they can show regulators they are statistically safer than human drivers that doesn't mean there won't be fatalities nor does it mean there won't be law suits but... As for joining a tesla network. I don't believe I'd use a $70K+ car to do that. I may pick up a stock M3 and use it. Too many people seem to enjoy keying teslas. If you let them inside alone...I can see real vandalism ahead.

 

[diplomat33]

As for what SAE level it is...who the F cares. If I can get into my car, tell it to go to Stop & Shop (a local grocery chain) and worst case all I have to do is put about 1 lb of torque on the steering wheel to signify I am paying attn, I'll call that FSD. If I travel from New England to NJ or MD or PA and it successfully navigates the George Washington Bridge Plaza I'll call that self driving too. It'll take regulators, be they state or if Congress preempts regulatory OK, to allpw Tesla to drop "Beta" nomenclature and drop the nags. SAE doesn't get a vote on that. SAE levels are great for those taking engineering classes.

First of all, the SAE levels are a convenient way of measuring how self-driving a car is and exactly what it can and cannot do. So the levels give us very useful information about the capabilities of the car. I would not be so quick to dismiss them.

In simple terms, the levels are:

Not self-driving:

Level 1: The car can perform steering OR acceleration/braking but the driver must supervise and monitor the environment

Level 2: The car can perform steering AND acceleration/braking but the driver must supervise and monitor the environment.

Self-driving:

Level 3: In certain conditions and environments, the car can perform all steering and acceleration/braking, can also monitor the environment but the car may request the "driver" intervene in certain situations where it might fail. The "driver" only needs to intervene when the car asks them to.

Level 4: In certain driving conditions and environments, the car can perform all steering and acceleration/braking, can also monitor the environment and can also intervene itself in cases where it will fail. The "driver" does not need to do anything at all when the car is in self-driving mode.

Level 5: the car can perform all steering and acceleration/braking, can also monitor the environment and can intervene in cases where it might fail in all conditions and environments. The "driver" never has to do anything at all.

Source: Defining Self-Driving Cars and Automated Vehicle Systems in SAE J 3016 - ANSI Blog

Full self-driving by definition means that you don't need to pay attention at all. So if you do need to pay attention, then it is not true self-driving. And if you are using a weight on the wheel, you are merely tricking the system into thinking that you are paying attention when in reality, you might not be. Fooling the system so that it does not nag you to hold the wheel, does not make the system full self-driving either.

Lastly, the ability of a car to drive from A to B without driver intervention, does not in itself necessarily make a car self-driving. That is actually a common misconception that many people have about self-driving cars.

What actually distinguishes a self-driving car from a non self-driving car is how much the car can monitor the environment and respond accordingly. For example, can the car track and avoid collisions with other objects such other vehicles or pedestrians, respond to traffic lights, traffic signs, etc... That's because it is one thing to automate driving where the car can steer and brake to get you from A to B, it is quite another to do it safely. In other words, it is pointless for a car to be able to drive from A to B if it can't keep its human passengers safe. True self-driving is when the car can get you from A to B safely and reliably every time such that you know you could read a book or watch a movie and you would be safe.

 

[wcorey]

Personally I'd prefer to have a car that allows me watch a movie or read a book...
There has been serious money on the table for quite a while now (I've read $80 billion+). I wouldn't hold my breath, it's an extremely difficult problem.

It was when they started. In theory some employees are already running their Teslas in FSD mode. It's been less than 2 yrs they've even mentioned trucks and semi cabs. Most people like you or me buy a Tesla for a multitude of reasons, conspicuous consumption, electric, cool, instant torque. FSD is lower on the list. Companies, they are buying fleets to replace drivers and do cross country in 2 1/2 days. Me, it was top of the list. In 2012 I bought my third Prius, paid extra for Advanced w/Adaptive CC. I could have gotten AS but not in that combination. I think it got 11 miles on battery. If I wanted 25 or AS I'd have to buy the next model year or the next... Their business model is selling cars as in repeat business. To buy a car that will improve via software...sure, I'm in. My Prius was still working fine except with all season tires I'd be lucky to see 9 miles. 310 is a big improvement.

 

[wcorey]

First of all, the SAE levels are a convenient way of measuring how self-driving a car is and exactly what it can and cannot do. So the levels give us very useful information about the capabilities of the car.

In simple terms, the levels are:

Not self-driving:

Level 1: The car can perform steering OR acceleration/braking but the driver must supervise and monitor the environment

Level 2: The car can perform steering AND acceleration/braking but the driver must supervise and monitor the environment.

Self-driving:

Level 3: In certain conditions and environments, the car can perform all steering and acceleration/braking, can also monitor the environment but the car may request the "driver" intervene in certain situations where it might fail. The "driver" only needs to intervene when the car asks them to.

Level 4: In certain driving conditions and environments, the car can perform all steering and acceleration/braking, can also monitor the environment and can also intervene itself in cases where it will fail. The "driver" does not need to do anything at all when the car is in self-driving mode.

Level 5: the car can perform all steering and acceleration/braking, can also monitor the environment and can intervene in cases where it might fail in all conditions and environments. The "driver" never has to do anything at all.

Source: Defining Self-Driving Cars and Automated Vehicle Systems in SAE J 3016 - ANSI Blog

Full self-driving by definition means that you don't need to pay attention at all. So if you do need to pay attention, then it is not true self-driving. And if you are using a weight on the wheel, you are merely tricking the system into thinking that you are paying attention when in reality, you might not be. Fooling the system so that it does not nag you to hold the wheel, does not make the system full self-driving either.

Lastly, the ability of a car to drive from A to B without driver intervention, does not in itself necessarily make a car self-driving. That is actually a common misconception that many people have about self-driving cars.

What actually distinguishes a self-driving car from a non self-driving car is how much the car can monitor the environment and respond accordingly. For example, can the car track and avoid collisions with other objects such other vehicles or pedestrians, respond to traffic lights, traffic signs, etc... That's because it is one thing to automate driving where the car can steer and brake to get you from A to B, it is quite another to do it safely. In other words, it is pointless for a car to be able to drive from A to B if it can't keep its human passengers safe. True self-driving is when the car can get you from A to B safely and reliably every time such that you know you could read a book or watch a movie and you would be safe.

I do get all that. As I said though, for the forseeable future if it can do all those tasks and all I have to do is make sure it doesn't attack a lane split barrier or ram an 18 wheeler broad side..I'll call it FSD for all intents and purposes. Do I prefer to sleep over a 12 hr trip? Yep but then there is that pesky every 2 1/2- 3 hr get up to recharge the battery. They don't appear to have the snake super chargers yet. Essentially, if there is an update to provide that functionality with the requirement to apply some torque, pay attention, and don't subvert the system come Oct or Nov...I would be thrilled. I'd expect as soon they aggregated several million miles in that mode with no consequential disengagements they'd drop the 'pay attn' requirement. They are the ones claiming robotaxis in 2020. If they weren't serious they'd let leasees buy out the car. Nope, they are stocking the Tesla Network.

 

[Daniel in SD]

It was when they started. In theory some employees are already running their Teslas in FSD mode. It's been less than 2 yrs they've even mentioned trucks and semi cabs. Most people like you or me buy a Tesla for a multitude of reasons, conspicuous consumption, electric, cool, instant torque. FSD is lower on the list. Companies, they are buying fleets to replace drivers and do cross country in 2 1/2 days. Me, it was top of the list. In 2012 I bought my third Prius, paid extra for Advanced w/Adaptive CC. I could have gotten AS but not in that combination. I think it got 11 miles on battery. If I wanted 25 or AS I'd have to buy the next model year or the next... Their business model is selling cars as in repeat business. To buy a car that will improve via software...sure, I'm in. My Prius was still working fine except with all season tires I'd be lucky to see 9 miles. 310 is a big improvement.

I guess I have to clarify. I'd like to be able to watch a movie or read a book with a probability of dying comparable to my own driving and also not face manslaughter charges if the car runs into someone.
I also didn't buy a Tesla for FSD (and I think there is nearly zero chance my car will ever drive itself).

 

[wcorey]

I guess I have to clarify. I'd like to be able to watch a movie or read a book with a probability of dying comparable to my own driving and also not face manslaughter charges if the car runs into someone.
I also didn't buy a Tesla for FSD (and I think there is nearly zero chance my car will ever drive itself).

Did you pay for FSD? If not, then you are correct, it'll never drive itself. If you added self driving I bet it would drive itself san dying or charged with manslaughter! That was pretty funny Dan...it got a chuckle.

There are people out there that firmly believe they are mere moments away from declaring bankruptcy.

 

[daniel]

Indulging you by searching for the original link.

ARK Contributes Its Financial Autonomous Taxi Model to GitHub

ps : Tesla EAP - ok. Zip files - definitely not

Thanks for that. I suspect that either their numbers or their analysis are off. Robo-taxis might indeed be cheaper than owning a car, since right now using Uber is probably cheaper than owning your own car. But people (myself included) enjoy the convenience of not having to call for a ride. Conventional taxis are not the best comparison, IMO, because the real competition to robo-taxis are Uber and Lyft. These are basically getting almost-free labor, using cars that don't have to have the hardware or software for FSD. So I don't see robo-taxis being any cheaper than Uber. In fact, I see robo-taxis being more expensive than Uber because I'd rather ride in a tobo-taxi than an Uber. It will be safer, therefore can command more money. Obviously, JMO.

Personally, I would love to have Level 3: Driver at the wheel but eyes off the road. I'd pay $5K for that in a heartbeat, and if Tesla wants to call it their FSD package, well, okay. But for me, "full self-driving" means I can take a nap in the back seat. So I don't consider anything below Level 4 to be truly FSD.

 

[electronblue]

There are people out there that firmly believe they are mere moments away from declaring bankruptcy.

Well if Elon is to be believed, they have been several times.

But look, given the recent past, it isn’t hard to imagine tons of reasons why AP2 cars might be unable to reach actual Level 4/5 self-driving (even with HW3 upgrade), even if you don’t think Tesla as a company is failing.

Tesla’s track-record on this has not been excatly confidence inspiring and it seems only prudent to assign some risk to them failing with the FSD goal (as marketed to us in 2016) on existing cars.

 

[electronblue]

And EAP has been feature complete for awhile now.

I have to do a rare partial disagree with @rnortman and a partial agree with @diplomat33 here, at least as far as the U.S. and EAP Design Studio description go.

Enhanced Autopilot seems to be feature complete at least in the U.S. I guess you could say it is not yet feature complete globally but I doubt that was the point here. You could also nitpick EAP uses eight not four cameras as originally announced but I doubt that was the point either.

One real thing EAP has not delivered on is AP1 parity though (speed sign reading) so that part of the original 2016 promise for EAP has not been fulfilled, but it was never in the Design Studio. A second thing is that until Enhanced Summon really ships (not just early access) the promised Summon maneuvers from Tesla’s Autopilot web page have not been completed but they were never in the Design Studio either.

Park Seek was never in the EAP Design Studio text, so at least the Design Studio text is indeed now feature complete for EAP in the U.S. I consider Park Seek as part of the original FSD promise and indeed it was part of the original 2016 FSD Design Studio text and FSD video.

For EAP the parking promise seems to have been just the regular AP1 type auto-parking.

 

[electronblue]

Different subject. There is way too much anonymity on here. Who on here is an automotive engineer? Who on here is actually working on self driving technology at Tesla? Being a little more blunt, who here knows what they are talking about? For myself, I'd like to know who I am talking to. To be sure, there are some who clearly know what they are talking about although I haven't the vaguest idea who they are. Asking for a friend.

I would assume nobody here is anything more than a Tesla past or current owner (or in some cases a prospective future Tesla owner or TSLA investor as the case may be). I certainly am nothing more than a Tesla car owner.

Isn’t that what owner’s forums like this are for? Researchind and sharing of information amongst peers. It is never perfect but then nothing in life is.

Even if we think Tesla knows what they are doing, we can’t trust them to tell us when it isn’t in their interests to be open about things. That is where peer information can be invaluable.

 

[wcorey]

But look, given the recent past, it isn’t hard to imagine tons of reasons why AP2 cars might be unable to reach actual Level 4/5 self-driving (even with HW3 upgrade), even if you don’t think Tesla as a company is failing.

Oh...I am totally comfortable with there is a point in the Tesla timeline below (before) which those cars, irrespective of what Elon promised, will never see level 4/5, unless the company takes the financial hit for retrofitting those cars to be AP2.5 compatible. Now, would Tesla survive that effort ? I'm not sure but pretty sure that question is why they won't attempt it.

 

[electronblue]

Oh...I am totally comfortable with there is a point in the Tesla timeline below (before) which those cars, irrespective of what Elon promised, will never see level 4/5, unless the company takes the financial hit for retrofitting those cars to be AP2.5 compatible. Now, would Tesla survive that effort ? I'm not sure but pretty sure that question is why they won't attempt it.

The bigger question may be, will even HW3 (as is shipping) achieve Level 4/5, or will it take some other suite.

 

[wcorey]

The bigger question may be, will even HW3 (as is shipping) achieve Level 4/5, or will it take some other suite.

Well, to be sure I don't know the answer to that. I believe it will. But I think (a point I was trying to make earlier) the answer hinges on what the definition of "human driver" is. Is it a 16 yr old on a learner's permit or a Nascar driver or trucker? Should anyone expect the level 5 (or 4) car to drive properly in a blinding snow storm? I would think not but it says 'human' driver so some drivers out there may be capable of driving in blinding snow (say Alaskan drivers) far better than someone from Austin TX. If you say, statistically the autonomous software must equal or exceed NTSB accident rates per million miles as well as fatalities per million miles driven. That is a quantitative benchmark.

Let's take airplanes. Pitot tubes are heated to a degree if touched it would more than burn. In commercial aircraft there are always more than one of them. Just in case. Is it reasonable to expect that level of protection around every camera and sensor on a car? Personally, I wouldn't expect it but then there are weather conditions I wouldn't drive in.

 

[electronblue]

@wcorey

Yep. Personally I am not convinced HW3 can read Level 4/5 even in fair weather but it is not solely about the suite itself. It is about the fact that Tesla’s progress in this area has been less than inspiring to me. So the risk in my mind has gone up somewhat.

 

[wcorey]

I would assume nobody here is anything more than a Tesla past or current owner (or in some cases a prospective future Tesla owner or TSLA investor as the case may be). I certainly am nothing more than a Tesla car owner.

Isn’t that what owner’s forums like this are for? Researchind and sharing of information amongst peers. It is never perfect but then nothing in life is.

Even if we think Tesla knows what they are doing, we can’t trust them to tell us when it isn’t in their interests to be open about things. That is where peer information can be invaluable.

Well, my point I was trying to make more delicately than perhaps necessary is if someone unfamiliar with software dev tries to BS about software development I can tell as that's what I did for a living. However, if someone tries to BS knowledge of neural networks I am unable to call BS as that is not a field I am knowledgeable in. I am on the cusp of deleting my Twitter account as it is truly a cesspool. The degree of civility is poor on a good day and the level of expert knowledge dissemination is incredibly difficult to judge unless the subject is something I am expert at. Where I just said 'I' I refer to anyone being able to judge. When identity is disclosed, it's easier to tell. It's more than just knowledge dissemination, it's also civility and demeanor. It's rare to find someone on here using their name, much less their location. So it raises a question. But, to each their own.

 

[daniel]

But look, given the recent past, it isn’t hard to imagine tons of reasons why AP2 cars might be unable to reach actual Level 4/5 self-driving (even with HW3 upgrade), even if you don’t think Tesla as a company is failing.

The bigger question may be, will even HW3 (as is shipping) achieve Level 4/5, or will it take some other suite.

In deference to wcorey I will state here that I have absolutely no competence in any of the relevant fields. That said...

Ever since Musk announced that the Model 3, as then being delivered, had all the hardware for FSD, which he described as being capable of driving with nobody in the car, I began saying that I thought he was mistaken. (His unreasonable optimism again, which is not a bad thing since it has driven him to accomplish things nobody else would try.) Specifically I believed that the sensor suite was inadequate. It turned out that the computer was inadequate. And since the software has not yet been developed I will assert that we don't yet know if the HW3 computer will be adequate or not. I still believe the sensor suite is inadequate.

I admire Elon Musk tremendously and I love my Tesla and I loved the Tesla I owned before this one. And if I could say one thing directly to him, it would be "Thank you for building my car!" But I do not believe that he can meet any of his self-imposed deadlines or that Tesla can turn the Model 3 into a robo-taxi with the hardware currently being delivered.

I think that Tesla will make a driverless car. I think it will have different hardware than is currently being put on the cars, and will take far longer than he has stated. They might, however, introduce some new feature(s) for the "FSD package" cars that they don't give to the EAP cars. I don't think they have yet. (?)

 

[electronblue]

In deference to wcorey I will state here that I have absolutely no competence in any of the relevant fields. That said...

Ever since Musk announced that the Model 3, as then being delivered, had all the hardware for FSD, which he described as being capable of driving with nobody in the car, I began saying that I thought he was mistaken. (His unreasonable optimism again, which is not a bad thing since it has driven him to accomplish things nobody else would try.) Specifically I believed that the sensor suite was inadequate. It turned out that the computer was inadequate. And since the software has not yet been developed I will assert that we don't yet know if the HW3 computer will be adequate or not. I still believe the sensor suite is inadequate.

I admire Elon Musk tremendously and I love my Tesla and I loved the Tesla I owned before this one. And if I could say one thing directly to him, it would be "Thank you for building my car!" But I do not believe that he can meet any of his self-imposed deadlines or that Tesla can turn the Model 3 into a robo-taxi with the hardware currently being delivered.

I think that Tesla will make a driverless car. I think it will have different hardware than is currently being put on the cars, and will take far longer than he has stated. They might, however, introduce some new feature(s) for the "FSD package" cars that they don't give to the EAP cars. I don't think they have yet. (?)

Your belief is one possible scenario, yes.

 

[timewasted]

With this talk about "Navigate on Autopilot is part one of Full Self Driving!", I find myself burying my face further and further into my hands. Every time I get an update, I turn on NoA to see if maybe this is the one that actually isn't terrible, and every time I sigh as I turn it back off. My commute involves a single freeway that is well marked and completely straight with relatively little traffic, and even with 2019.20.4.2 my car handles it as well as a timid teenager who is driving on the freeway for the first time. Micro-braking every time a car dares to be in the lane beside me, harder braking every time it nears an overpass, opting to cruise along well under the speed limit instead of moving to the faster open lane to its left, keeping the turn signal on without any action because there is a car somewhere behind/beside me. And that's all just from my morning commute today!

Seriously, is there anyone in the Las Vegas area that has an S/X that is actually happy with NoA? I'd LOVE to go for a ride in your car if so!

My apologies for my rant breaking up the discussion about how my S is going to become a level 5 robotaxi that operates everywhere without geofencing within the next 3 months maybe but 6 months definitely.