PSA: Please don't expect any FSD features this year, or even early next year, really

[MSR]

Other than removing the "confirm" requirement for auto lane changes, which they've said is coming very soon, EAP is feature complete already

How can you say that when AP 2.5 doesn't recognize speed limit signs, and in many cases has no idea what the actual speed limit is on many stretches of divided highway? I drive both an AP1 Model S and a Model 3 and can assure you that this is one area AP1 still outperforms AP 2.5.

Certainly any form of FSD will require a knowledge of the actual speed limit in addition to stop sign / stop light recognition & reaction at least as good as the average driver. Then there's the multitude of special cases ....

IMO, FSD requiring supervision is an oxymoron.

 

[Knightshade]

How can you say that when AP 2.5 doesn't recognize speed limit signs, and in many cases has no idea what the actual speed limit is on many stretches of divided highway?

Because speed limit recognition is not a feature of EAP.

And never has been.

And never was promised to be.

So...I can say that pretty easily.

Especially when I just quoted to you the features EAP actually was promised to contain. And it already contains all of them, today, other than changing lanes without waiting for approval (which they've promised is coming very soon).

I drive both an AP1 Model S and a Model 3 and can assure you that this is one area AP1 still outperforms AP 2.5.

That's great. But AP and EAP aren't the same thing.

Certainly any form of FSD will require a knowledge of the actual speed limit in addition to stop sign / stop light recognition & reaction at least as good as the average driver.

Sure. But FSD also is not EAP.

I fully expect FSD to read speed limits.

Not only that- If Tesla is at all smart, they'll then use that FSD speed data to update the database they use for HW2.x EAP cars.

Thus improving EAP by leveraging the more advanced systems in the HW3 cars.

 

[MSR]

Because speed limit recognition is not a feature of EAP.

And never has been.

And never was promised to be.

So...I can say that pretty easily.

Especially when I just quoted to you the features EAP actually was promised to contain. And it already contains all of them, today, other than changing lanes without waiting for approval (which they've promised is coming very soon).


....

Really? So it's your opinion Tesla's promise that "Your Tesla will match speed to traffic conditions" can be done without knowledge of the speed limit.

Not to mention the fact that the E in EAP stands for Enhanced. So it's ok by you if EAP costing twice as much ($5K) as AP ($2.5K) never includes all the features of AP.

Sorry, I guess we'll just have to agree to disagree. I sincerely hope Tesla doesn't think like you and includes all the features that are possible in EAP to increase safety.

 

[Knightshade]

Really? So it's your opinion Tesla's promise that "Your Tesla will match speed to traffic conditions" can be done without knowledge of the speed limit.

Of course.

That's TACC. Which works great.

The posted limit has nothing to do with 'traffic conditions'

Not to mention the fact that the E in EAP stands for Enhanced. So it's ok by you if EAP costing twice as much ($5K) as AP ($2.5K) never includes all the features of AP.

Sure, since it includes tons of features AP does not. Nav on AP for example (which includes a number of capabilities AP1 can't do).

Sorry, I guess we'll just have to agree to disagree. I sincerely hope Tesla doesn't think like you and includes all the features that are possible in EAP to increase safety.

Safety features are free (blind spot warning, emergency braking, etc)- they're not part of AP or EAP- though again EAP cars will generally be a lot better at them than AP1 cars due to more, better, sensors.


Driver assist features like EAP, which, again, has a very clearly defined feature set and has for years, and is feature complete with the one exception I mentioned, are not free.

Likewise FSD features, which will be everything beyond the already-known EAP features, won't be free either.


None of this is really an "opinion" you can disagree with- Tesla has been pretty clear, in writing, what EAP is and is not, since it launched.

 

[StealthP3D]

And Tesla is responsible for accidents the occur while the system is being tested. How fast would the DMV pull their permit if there were multiple major accidents?

There will always be multiple major accidents whether computers are driving or humans. What is significant is the rate of accidents per million miles and the severity of them.

I'm confident that FSD won't be released for use without human intervention until it can be safer than humans by a factor of 2.

The average human is pretty undependable in this role so I think we are getting somewhat close to having self-driving cars that are safer than the average human.

 

[Daniel in SD]

There will always be multiple major accidents whether computers are driving or humans. What is significant is the rate of accidents per million miles and the severity of them.

I'm confident that FSD won't be released for use without human intervention until it can be safer than humans by a factor of 2.

The average human is pretty undependable in this role so I think we are getting somewhat close to having self-driving cars that are safer than the average human.

Obviously any FSD system will have accidents. My point was that current regulations require the manufacturer to be liable for them. Tesla's permit for testing FSD in California requires a test driver who is trained and ready to take over at any time.
Tesla seems to want to have regulations changed to allow untrained customers test FSD and have them be liable for accidents that occur while the system is in use. I don't think that regulators will allow that...

 

[Knightshade]

Obviously any FSD system will have accidents. My point was that current regulations require the manufacturer to be liable for them.

Your point is factually not correct though.

For L3 and higher each state differs. Some have no laws on this at all. Some have laws that work the way you suggest. Some have laws the explicitly DO NOT work that way.

In Nebraska for example it's already legal right now to operate a driverless, self-driving, car as long as it can obey all applicable driving/safety laws and can achieve a minimal risk condition if the system malfunctions (basically being able to pull over safely if there's a problem).

And it also requires you submit proof to the DMV that the vehicle has insurance that meets state requirements. Which would be the responsibility of the owner/operator of the car, not the people who made the car (unless that's who is operating the thing)


Quite a few states actually work this way... the owner/operator is responsible, not the manufacturer.

I can find more that work this way than who consider the manufacturer responsible for the system in fact.

Tesla's permit for testing FSD in California requires a test driver who is trained and ready to take over at any time.
Tesla seems to want to have regulations changed to allow untrained customers test FSD and have them be liable for accidents that occur while the system is in use. I don't think that regulators will allow that...

That's not what is being described- it's being suggested Tesla would release new FSD-only features as L2. In which case the driver is always responsible, just as with EAP (an L2 system) today.

 

[S4WRXTTCS]

Musk's argument is that LIDAR is not needed because humans don't have LIDAR, they just have two eyes. So when the Model 3 has a human decision making capability then Musk's argument holds water, until then, augmenting with additional hardware sensors seems like a prudent choice and Tesla themselves are doing it via the use of forward radar and ultrasonic sensors, all of which augment the "eyes" that the car has.

The funny thing is quite a few drivers use the cars system to augment our own visual system.

I'm sure I'm not the only one who has used the HW2.5 radar+camera output to lower my stress during limited visibility moments. If they added some fairly weather immune Lidar to that it would be awesome.

They don't need the entire 360 scanning lidar. But, maybe the newer/cheaper solid state lidar.

Lidar would also immensely help with Tesla's struggle between too many false positives in AP/TACC, and too many false negatives. The camera and the radar would be augmented by the Lidar sensor to help with redundancy.

When Elon was against Lidar it made sense, but I don't think it makes any sense any more.

I also doubt it's really going to up to Elon. Other manufactures of autonomous vehicles will force Lidar into specifications required for L4 driving, and above.

 

[voip-ninja]

The funny thing is quite a few drivers use the cars system to augment our own visual system.

I'm sure I'm not the only one who has used the HW2.5 radar+camera output to lower my stress during limited visibility moments. If they added some fairly weather immune Lidar to that it would be awesome.

They don't need the entire 360 scanning lidar. But, maybe the newer/cheaper solid state lidar.

Lidar would also immensely help with Tesla's struggle between too many false positives in AP/TACC, and too many false negatives. The camera and the radar would be augmented by the Lidar sensor to help with redundancy.

When Elon was against Lidar it made sense, but I don't think it makes any sense any more.

I also doubt it's really going to up to Elon. Other manufactures of autonomous vehicles will force Lidar into specifications required for L4 driving, and above.

I agree wholeheartedly with most of what you say but I don't know that Lidar as a technology will be required if other vendors can prove that they are able to execute without it.

Where Lidar provides superior capabilities to visual/radar/ultrasonic sensors is in situations where the visual cameras, much like a human, aren't sure of what they are looking at.

Scanning ahead and seeing a pothole that could damage the car is trivial for LIDAR but is not easy for a camera system to figure out. Even humans with thousands of hours of driving time have a hard time making this determination.

Another example I can think of is the number of times I have approached a stop sign where the sign is partially obstructed by overgrowth. With visual sensors it is possible that the car is not going to see enough of the sign and see it clearly enough to know it's a stop sign. With LIDAR the sign can be scanned at high speed from a good distance away and the car can then match that partial image and know immediately that it's SUPPOSED to be a stop sign even if visually it's obstructed to some degree.

Another area where LIDAR offers some real advantage is in knowing what is "normal" in a particular area and comparing that known map with what the LIDAR is scanning. This super high resolution map that can be compared with what the LIDAR sees might ultimately be what is needed to safely navigate complex urban areas with narrow roads, convoluted entrances, exits, alleyways, etc.

 

[S4WRXTTCS]

In Nebraska for example it's already legal right now to operate a driverless, self-driving, car as long as it can obey all applicable driving/safety laws and can achieve a minimal risk condition if the system malfunctions (basically being able to pull over safely if there's a problem).

And it also requires you submit proof to the DMV that the vehicle has insurance that meets state requirements. Which would be the responsibility of the owner/operator of the car, not the people who made the car (unless that's who is operating the thing)

Quite a few states actually work this way... the owner/operator is responsible, not the manufacturer.

Related to this is insurance companies are already changing their policies to say that they're not going to cover an accident if the car was in a self-driving mode.

So effectively you can't do L3 driving in a vehicle unless either your insurance company is willing to cover it or the Manufacture is willing to cover it.

In that case I'd get a Volvo because Volvo says they'll cover it (L3 and higher).

The Tesla solution will likely be Tesla insurance.

I really wish the Audi A8 with L3 capability was being released in the States because I'd love to see how the insurance thing will work out on that.

 

[Daniel in SD]

In Nebraska for example it's already legal right now to operate a driverless, self-driving, car as long as it can obey all applicable driving/safety laws and can achieve a minimal risk condition if the system malfunctions (basically being able to pull over safely if there's a problem).

Don't have time to find whole text of law but:

On Tuesday, a state senator attempted to clarify that question about the fast-growing technology.

Under a legislative bill introduced by State Sen. Suzanne Geist of Lincoln, the manufacturer of the self-driving technology would be liable in a crash involving a vehicle that was capable of “the entire driving task” and operating in autonomous mode.

Who's liable when a self-driving car runs you over? Legislative bill attempts to clarify that

 

[S4WRXTTCS]

I agree wholeheartedly with most of what you say but I don't know that Lidar as a technology will be required if other vendors can prove that they are able to execute without it.

Where Lidar provides superior capabilities to visual/radar/ultrasonic sensors is in situations where the visual cameras, much like a human, aren't sure of what they are looking at.

Scanning ahead and seeing a pothole that could damage the car is trivial for LIDAR but is not easy for a camera system to figure out. Even humans with thousands of hours of driving time have a hard time making this determination.

Another example I can think of is the number of times I have approached a stop sign where the sign is partially obstructed by overgrowth. With visual sensors it is possible that the car is not going to see enough of the sign and see it clearly enough to know it's a stop sign. With LIDAR the sign can be scanned at high speed from a good distance away and the car can then match that partial image and know immediately that it's SUPPOSED to be a stop sign even if visually it's obstructed to some degree.

Another area where LIDAR offers some real advantage is in knowing what is "normal" in a particular area and comparing that known map with what the LIDAR is scanning. This super high resolution map that can be compared with what the LIDAR sees might ultimately be what is needed to safely navigate complex urban areas with narrow roads, convoluted entrances, exits, alleyways, etc.

Doesn't the L3 capable Audi A8 have Lidar?

How do the semi-autonomous systems in the new Audi A8 stack up?

It says Laser Scanner on that.

 

[Jonathantuba]

The first FSD feature, Navigate on Autopilot is in effect already out. I think once AP3 is installed, more features will be rolled out to cars capable (or upgraded with AP3) and during 2020 FSD on main highways will become technically possible, although if legally allowed remains to be seen. I think on that sort of road Tesla is already 95% of the way there.

FSD on other roads will take much longer. Personally I think sleeping from home to destination is several years away. But if I can leave the car to drive on main highways on long journeys, so I can work or rest I will be well satisfied and think paying for FSD worthwhile.

 

[Knightshade]

Don't have time to find whole text of law but:


Who's liable when a self-driving car runs you over? Legislative bill attempts to clarify that

So first- that very link contradicts your claims the MFG is currently responsible....so thanks for confirming your original claim was incorrect

Second your link is about a proposed law.

My remark was about the actual current law in Nebraska.

Here's the current, actual, law... passed just last year in fact-
https://nebraskalegislature.gov/FloorDocs/105/PDF/Slip/LB989.pdf

Before an automated-driving-system-equipped vehicle may operate on the public roads of this state, a person shall submit proof of financial responsibility satisfactory to the department that the automated-driving system-equipped vehicle is covered by insurance or proof of self-insurance that satisfies the requirements of the Motor Vehicle Safety Responsibility Act

So the new law isn't so much "clarifying" who is liable- the current law already does that. It's trying to CHANGE who is responsible.

Currently it ain't the manufacturer (unless they are the ones owning/operating the car).

And that's the case in most states that have any laws on the topic from what I can tell.

Related to this is insurance companies are already changing their policies to say that they're not going to cover an accident if the car was in a self-driving mode.

So effectively you can't do L3 driving in a vehicle unless either your insurance company is willing to cover it or the Manufacture is willing to cover it.

Nope. Nebraska (and many states) have a third option. Posting a bond as insurance of yourself (or your car in the case of a self driving car).

It's not cheap, but it's an option.

 

[Knightshade]

The first FSD feature, Navigate on Autopilot is in effect already out

Nope. That's an EAP feature. Explicitly.

It is (once they remove the confirm-lane-change requirement) the last piece of functionality originally promised as part of EAP in fact.


Certainly FSD is likely to offer an extended-functionality version of such a feature as part of FSD, once it gets good enough on non-divided highways, but that's a different thing (current Nav on AP is specifically for such highways as part of EAP).

 

[voip-ninja]

Doesn't the L3 capable Audi A8 have Lidar?

How do the semi-autonomous systems in the new Audi A8 stack up?

It says Laser Scanner on that.

Laser Scanner definitely is LIDAR.

 

[Elmexiken]

I think the OP's expectations of "feature complete " fsd and actually implementation and deployment are not in alignment with reality

Definition of feature complete from wikipedia "A feature complete version of a piece of software has all of its planned or primaryfeatures implemented but is not yet final due to bugs, performance or stability issues. This occurs at the end of alpha testing ofdevelopment."

An Alpha release is a long way from beta, external user tests or even final candidates, let alone when state regulatory bodies may approve such technology.

The reality is, we are probably a few years off from that tech working in our cars and the software and hardware required could change substantially.

I would love to see Elon prove me wrong though.

No, this is exactly what I'm saying. I believe feature complete means almost little to nothing, for us, just as it doesnt when building a video game, for example. I don't see any features launching until well into 2020.

 

[Knightshade]

No, this is exactly what I'm saying. I believe feature complete means almost little to nothing, for us, just as it doesnt when building a video game, for example. I don't see any features launching until well into 2020.

So you think Elon was lying when he explicitly said the first FSD only features would launch same time as HW3 hardware, which is coming early this year?

 

[davedavedave]

1. This is self-evidently Musk's belated gambit...
2.... 3...4...5....
6. Taken together this means I think it ~55% probable...

I was really trying to follow, but then you *really* lost me at 55%

 

[Daniel in SD]

So the new law isn't so much "clarifying" who is liable- the current law already does that. It's trying to CHANGE who is responsible.

I would say that the law is vague and that is why it needs clarifying. It seems very likely that someone could sue the manufacturer of the self driving system and win in court. Is there any state that explicitly exempts the manufacturer from liability? That certainly doesn't seem like the way the law is heading. Traditionally manufacturers have been held liable for defects in their designs.