What will happen within the next 6 1/2 weeks?

[am_dmd]

Tesla CSR is now trying to spin off FSD as level 3, 4 or 5. In short moving the goal post and changing FSD definitions as they see fit. I would be happy to pay for whatever Telsa wants to charge me if Musk puts one of kid in the back seat and let the car do what that fake FSD video did ¯\_(ツ)_/¯

 

[_jal_]

Strike "almost" from the above. "Feature complete" is exactly an internal engineering benchmark and means nothing to drivers, though it might be interesting to fans of the process who enjoy the story of how an engineering project progresses

Then I wish he’d stop talking in terms like that in public. I’m not in software, and to be fair to myself, from Elon’s comment I’d have no reason to be aware that “feature complete” is software development jargon. I didn’t believe it because I’ve been reading this stuff long enough, but as a reasonable consumer I’d assume a feature-complete full self driving system to be a full self driving system that has a feature set that’s complete as defined. And not defined just in abstract terms but with a feature that allows automatic driving on city streets before the end of the year. (That’s on the website right now btw)

All the semantic arguments about what “is” is sort of stick in my craw because the mass of people are going to take these terms and words prima facia as a common sense description of features. If “full” means something technical or “complete” is an industry adjective, then there should be a blog post at the very least. It’s disingenuous at best and fraudulent at worst to take $7,000 from people for a system that specs automatic driving on city streets as a feature that cannot actually automatically drive on city streets. And going in a straight line, stopping for obstacles and recognizing stop signs and stop lights does not in any common sense interpretation actually constitute automatically driving on city streets

 

[diplomat33]

Well, not really where we differ mainly, the geofence thing is more of a side point. I think the notion that Tesla is any kind of feature complete for Level 5 (within U.S. or any market for that matter) is pretty far off.

But that is their expectation for end of 2019 and they haven’t retracted it, so I guess the claim still stands. Level 5 no geofence feature complete by end of 2019...

We are just a few days away from Jan 1, 2020. I think at this point, we can be certain that Tesla will miss their end of the year deadline for "feature complete".

I regard L5 as being "No driver needed, ever." "FSD" means the same thing to me, though apparently not to everyone. To me, a self-driving car is one that drives itself. A full self-driving car is one that is always self-driving. My suggested acronym: NNaD: (Never Needs a Driver.)

Yes, a self-driving car is one that drives itself. But self-driving just means that while the system is active, the car is doing 100% of the driving. That is one reason why the SAE has levels and talks about the ODD because you still need to define when and where the self-driving can be turned on.

L3 is true self-driving when the system is on (ie car does all the driving) but you can't use it necessarily everywhere or anytime and the driver is still needed as the fallback.

L4 is true self-driving when the system is on and you still can't use it everywhere or anytime. However, when it is on, the driver is not needed.

L5 is true self-driving when it is on and it can be used anytime and everywhere. The driver is only needed if they wish to drive manually.

 

[DanCar]

We are just a few days away from Jan 1, 2020. I think at this point, we can be certain that Tesla will miss their end of the year deadline for "feature complete". ...

Later Elon said at least for early access (EA) peeps, and there has been a rumor or two that some EA people have it. Sometimes people have EA for 4 months before general release, so that gives a possible timeline before everyone gets it, April.

 

[diplomat33]

Later Elon said at least for early access (EA) peeps, and there has been a rumor or two that some EA people have it. Sometimes people have EA for 4 months before general release, so that gives a possible timeline before everyone gets it, April.

Yes, I am aware of this. However, the website stated that the public would get both traffic light/stop sign response and automatic driving on city streets by the end of this year. That will not happen. So, I think Elon definitely moved the goalposts when he saw that they would not be ready to deliver those features to the public by the deadline.

 

[SouthSeas]

We should stop lying to ourselves and call it what it is, change FSD to PSD. Partial self driving.
Elon said HW4 will be 3 times better than HW3.

I'm not sure 3X will allow a big improvement in features. The sort of processing required will be looking for 10X improvements in performance to gain much in the way of capabilities.

 

[SouthSeas]

Yes, I am aware of this. However, the website stated that the public would get both traffic light/stop sign response and automatic driving on city streets by the end of this year. That will not happen. So, I think Elon definitely moved the goalposts when he saw that they would not be ready to deliver those features to the public by the deadline.

When did Elon ever mention any goal posts that weren't on the bed of an F150?

 

[DanCar]

Yes, I am aware of this. However, the website stated that the public would get both traffic light/stop sign response and automatic driving on city streets by the end of this year. That will not happen. So, I think Elon definitely moved the goalposts when he saw that they would not be ready to deliver those features to the public by the deadline.

Yeah, setting impossible deadlines is an Elon hallmark.

 

[DanCar]

I'm not sure 3X will allow a big improvement in features. The sort of processing required will be looking for 10X improvements in performance to gain much in the way of capabilities.

The assumption in this argument is that when Elon said 3x better he meant 3 times more processing power. One could instead assume he literally meant three times better, and therefore all features are 3 times better and or there are 3 times as many features.

 

[DanCar]

... However, the website stated that the public would get both traffic light/stop sign response and automatic driving on city streets by the end of this year. ...

Link? I found what the current AP page says interesting: Autopilot

The future use of these features without supervision is dependent on achieving reliability far in excess of human drivers as demonstrated by billions of miles of experience ...

It is interesting because Tesla is not stating a timeline. So if it happens in 2040, oh well.

 

[diplomat33]

Yeah, setting impossible deadlines is an Elon hallmark.

I wish he didn't. But I guess asking Elon to stop making impossible deadlines would be like asking the sun to stop rising in the morning.

The assumption in this argument is that when Elon said 3x better he meant 3 times more processing power. One could also assume he literally meant three times better, and therefore all features are 3 times better and or there 3 times as many features.

3x more processing power would make sense because that is a quantity that can be measured. I am not sure what 3x better would mean. How would you measure that?

Link? I found what the current AP page says interesting: Autopilot

Not the autopilot website. I was referring to the order page when you order a new car. When you go to the autopilot section for ordering a new car, you can add the FSD option. It still lists traffic light/stop sign response and automatic driving on city streets as FSD features that are coming later this year. So the order page does state a timeline.

 

[S4WRXTTCS]

That is not correct. If someone posted such a thing, it would be wrong. Well, misleading at least.

Level 5 is allowed to be limited to a market (say, the U.S.) and human driveable conditions but limiting it to some other ODD would demote it to Level 4. So it can have a geofence (to the market) and refuse to work in undriveable conditions (as long as it can reach MRC) but not really a limited ODD.

This is precisely why I don't like the way Level 5 is defined.

It makes an assumption that there is a defined ODD for humans, and that's not the case. What most humans have is an adaptable ODD that depends on how much risk they're willing to take. A robo-driver can't adjust it's ODD just because some human needs to get from Point A to Point B in a snow storm.

The other problem that I have with that level is it doesn't allow the market place to be competitive. If the ODD is greater for one vehicle make/model than another then that needs to be clearly communicated to customers.

We already have the limited ODD problem when it comes to Uber/Lyft/etc. That the ODD that most human drivers are assumed to have isn't the same ODD that they're actually willing to risk while giving a ride share. The last time the PNW had a snowstorm I couldn't get an Uber/Lyft for a friend to go to the airport. I ended up having to drive her to the closest light rail station.

What I like about L4 is it gives room to grow an ODD, and doesn't make comparisons to human drivers.

L5 should be similar in that the ODD should grow beyond that of what defined it as Level 5.

It shouldn't be left to the manufacture to define the ODD, but there should be an industry standard of what's expected to go beyond the first level of L5. Where things like snow rating was added, speed rating. and if it's able to see through fog with special sensors.

 

[daniel]

Yeah, setting impossible deadlines is an Elon hallmark.

Which is the same personality trait that has given us the company that made these fabulous cars. A person who was not a super-optimist never would have produced the things Musk has.

 

[S4WRXTTCS]

A person who was not a super-optimist never would have produced the things Musk has.

Or become president of the USA with zero political experience.

Edit:

This wasn't meant to be political. It was simply a joke about the power of optimism in the world today. That in the reality we live in today there are a lot of examples of people having success where unwavering optimism played a large role.

In 2016 we had a company selling FSD despite no one ever achieving even L4 driving over a large non-geofenced area.

In 2016 we also elected a president who had no prior political experience.

Regardless of how one thinks about either it's hard to deny the role super optimism played.

 

[diplomat33]

It makes an assumption that there is a defined ODD for humans, and that's not the case.

There is certainly a maximum ODD for human drivers. It's all conditions where the road is navigable (ie not flooded or iced over) and visibility is good enough to see.

L5 should be similar in that the ODD should grow beyond that of what defined it as Level 5.

But L5 ODD already encompasses all driver manageable road conditions with no restrictions whatsoever. The only ODD that would be greater than L5 ODD would be one that involves non driver manageable conditions.

L5 ODD:

"“Unconditional/not ODD-specific” means that the ADS can operate the vehicle under all driver-manageable road conditions within its region of the world. 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)."

 

[EVNow]

We are just a few days away from Jan 1, 2020. I think at this point, we can be certain that Tesla will miss their end of the year deadline for "feature complete".

Musk stated this clearly in a tweet.

Needs a few more days of validation, then early access, then wide release.​

 

[S4WRXTTCS]

There is certainly a maximum ODD for human drivers. It's all conditions where the road is navigable (ie not flooded or iced over) and visibility is good enough to see.



But L5 ODD already encompasses all driver manageable road conditions with no restrictions whatsoever. The only ODD that would be greater than L5 ODD would be one that involves non driver manageable conditions.

L5 ODD:

"“Unconditional/not ODD-specific” means that the ADS can operate the vehicle under all driver-manageable road conditions within its region of the world. 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)."

You seem to think that the ODD for humans is a binary thing, but that's not the case. There are a great many situations that impact whether one human continuous on with a journey compared to another.

Sometimes the people that get off the road have anxiety with whatever they're being faced with. Like the first time I drove my camper van I decided it was best to stop for the night when I saw a lighting storm ahead, and the wind was throwing me around. The road was navigable, and the conditions didn't dictate that I got off the road immediately. I chose to do so mostly because I felt like things were getting worse.

There have been plenty of other times where I drove in bad situations, but only because of more familiarity with the situation at hand.

Lots of times its not a binary case as usually when that happens the roads themselves get closed. But, it's simply a judgement call based on the information we have at the time. Sometimes its based on prior knowledge like if I know a bridge typically freezes over in such and such conditions. So I simply won't drive on the bridge in these conditions, and later on I'll read in the news that 40-50 cars had some pileup on that bridge due to it.

In the camper van case it's likely an L5 system would have continued on assessing the situation not based on what it thought human drivers would do, but based on a design element. Like if estimated cross wind speeds are greater than this then we shouldn't continue.

In the other cases I likely exceeded the ODD parameters that would be practical for an initial L5 car.

Short term my expectation from L5 (when they get released) is they'll have ODD limits that seem overly careful.

Long term my expectation from a L5 vehicle is that they'll exceed the ODD capabilities of even the best human drivers because they'll have better sensor equipment. Like instruments that can see through fog, and more connectively to what's happening with tires.

 

[diplomat33]

You seem to think that the ODD for humans is a binary thing, but that's not the case. There are a great many situations that impact whether one human continuous on with a journey compared to another.

Sometimes the people that get off the road have anxiety with whatever they're being faced with. Like the first time I drove my camper van I decided it was best to stop for the night when I saw a lighting storm ahead, and the wind was throwing me around. The road was navigable, and the conditions didn't dictate that I got off the road immediately. I chose to do so mostly because I felt like things were getting worse.

There have been plenty of other times where I drove in bad situations, but only because of more familiarity with the situation at hand.

Lots of times its not a binary case as usually when that happens the roads themselves get closed. But, it's simply a judgement call based on the information we have at the time. Sometimes its based on prior knowledge like if I know a bridge typically freezes over in such and such conditions. So I simply won't drive on the bridge in these conditions, and later on I'll read in the news that 40-50 cars had some pileup on that bridge due to it.

In the camper van case it's likely an L5 system would have continued on assessing the situation not based on what it thought human drivers would do, but based on a design element. Like if estimated cross wind speeds are greater than this then we shouldn't continue.

In the other cases I likely exceeded the ODD parameters that would be practical for an initial L5 car.

Short term my expectation from L5 (when they get released) is they'll have ODD limits that seem overly careful.

Long term my expectation from a L5 vehicle is that they'll exceed the ODD capabilities of even the best human drivers because they'll have better sensor equipment. Like instruments that can see through fog, and more connectively to what's happening with tires.

You see to think that the ODD is fluid and basically changes hour by hour and driver by driver. And maybe human drivers misjudge their own ODD. But My point is that ODD, especially for an autonomous car, needs to be fixed. For a L5 car, I don't think it can fluid.And if the conditions look ok but get worse, an L5 car will need to know when to pull over as soon as the conditions cross over into outside it's ODD.

 

[electronblue]

We are just a few days away from Jan 1, 2020. I think at this point, we can be certain that Tesla will miss their end of the year deadline for "feature complete".

I’m not so concerned whether or not Tesla misses ”feature complete” at the end of 2019, after all it was always an internal deadline that would not be visible to the outside anyway.

What I am concerned is that their ”feature complete” may not have anything to do with ”Level 5 no geofence feature complete” which was Tesla’s claim on Autonomy Investor Day.

 

[electronblue]

Which is the same personality trait that has given us the company that made these fabulous cars. A person who was not a super-optimist never would have produced the things Musk has.

I think it can be argued Tesla would have done just as well or even better had they not made the lofty ”sleep in your car” Level 5 no geofence ”feature complete” promises for AP2/3 in 2016-2019 and just focused on great BEV technology and gradual AP progress.

I continue to think this may be one area when Tesla unnecessarily shot themselves in the foot, similar to the taking private debacle. Now, that doesn’t mean they can’t survive it (yes they can), but these can still be unnecessary self-inflicted wounds.