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

[S4WRXTTCS]

Oh I’m sure there will be test drives. I do not doubt that. The video mention did not mean to exclude test drives. What level of difficulty or freedom they offer remains to be seen.

After watching the enhanced summons "test drives" I think we can safely rule out parking lots.

I expect that they won't allow autonomous operation in parking lots. If they did that alone would show significant progress since we already have a baseline of expectation from enhanced summons.

 

[electronblue]

I think the Tesla Network is purely something for investors that actual owners like myself or you should ignore.

You can blame Elon/Tesla for it, or you can blame the way the stock market works. Where they value things not on what you can do today, but what you expect to do tomorrow.

I do expect the Tesla network to eventually be a thing, but not for another 3-5 years.

I'm fully on board with the idea that Tesla will make gradual progress, but that progress is going to be significantly more visible than what it has been.

There is still going to be lots of problem points. Ridiculous things that shouldn't happen.

For HW3 I expect a pretty significant gain in both accuracy and capabilities (what it can detect) almost out of the gate (within 3-6 months of the start of releases). What they'll release will be so significant to actual drivers that we won't even care that it doesn't hit the mark of actual FSD.

I expect to see Tesla to completely drop the idea of trying to upgrade cars. They'll upgrade AP2/AP2.5 to HW3, but they won't make that mistake again.

So we'll see HW4, HW5, etc. Where each things brings us closer to L5 FSD.

I expect the entire automotive market place to converge where there are multiple L3 offering right around the same time. I don't expect one company to be that far ahead of another when it comes to autonomous. If there is there is likely some huge caveat to it like it being fleet vehicles only (Waymo).

Your thesis sounds quite close to my second thesis: Tesla is bluffing (but there will also be gradual progress).

The only thing really where we seem to diverge is the Level 3 part. I have a really hard time seeing Tesla reach car responsible driving with their camera only suite in the same timeframe as others can be expected to. Tesla has a major deployment and cohones advantage but that by definition tends to given them more in the Level 2 department than in Level 3... (Unless Tesla isn’t bluffing and there is a major leap ahead.)

I am not dismissing any possibility, just trying to spell out my internal thought process.

 

[CarlK]

After watching the enhanced summons "test drives" I think we can safely rule out parking lots.

I expect that they won't allow autonomous operation in parking lots. If they did that alone would show significant progress since we already have a baseline of expectation from enhanced summons.

Traffic laws do not apply in privately owned parking lots or driveways. Police will not issue you a ticket for not obeying the stop sign, for example, although that will not relieve you of responsibility for an accident. I think that's the reason why summon is the first FSD feature to be released.

 

[S4WRXTTCS]

Traffic laws do not apply in privately owned parking lots or driveways. Police will not issue you a ticket for not obeying the stop sign, for example, although that will not relieve you of responsibility for an accident. I think that's the reason why summon is the first FSD feature to be released.

Oh, crap.

You summoned Blader by saying FSD feature.

 

[electronblue]

Traffic laws do not apply in privately owned parking lots or driveways. Police will not issue you a ticket for not obeying the stop sign, for example, although that will not relieve you of responsibility for an accident. I think that's the reason why summon is the first FSD feature to be released.

To be clear though smart Summon is originally an EAP feature so no FSD only features in the original sense have been released yet. (Some might argue smart Summon is an AP1 feature since it was promised already back then. )

 

[CarlK]

To be clear though smart Summon is originally an EAP feature so no FSD only features in the original sense have been released yet. (Some might argue smart Summon is an AP1 feature since it was promised already back then. )

I'm very clear of what I meant by FSD and why smart summon does not need to wait for regulatory approval. Couldn't care less what name you want to call it. Tired of nitpicking just for argument's sake.

Oh, crap.

You summoned Blader by saying FSD feature.

Wasn't involved in those conversations but couldn't care even less of what Blader says.

 

[electronblue]

I think that's the reason why summon is the first FSD feature to be released.

I'm very clear of what I meant by FSD and why smart summon does not need to wait for regulatory approval. Couldn't care less what name you want to call it. Tired of nitpicking just for argument's sake.

I believe you know full well why the ”first FSD feature” is an important distinction though. I think accuracy there helps because all old FSD owners here are still waiting for that first FSD feature to come.

Smart Summon is an EAP feature so not that first FSD feature.

 

[Bladerskb]

And you were wrong about NOA without confirmation, so don't talk to me about embarrassing myself, please.

How many times do i have to be right?

 

[electronblue]

If Elon is right it would seem to be looking for a torque force then — and there not just an existing torque level sometime prior?

Hard one... to believe Elon now or not...

 

[mongo]

FFS GIFy, that is the description of the nag system.

It is verifying a torque to confirm hands on wheel (level triggered), not looking for an additional torque to confirm the lane change (edge triggered)...
Next tweet (currently)

 

[electronblue]

FFS GIFy, that is the description of the nag system.

It is verifying a torque to confirm hands on wheel (level triggered), not looking for an additional torque to confirm the lane change (edge triggered)...
Next tweet (currently)
View attachment 394898

Well guess it depends.

Say the torque systems looks for that torque every 10 seconds..

Checkpoint - wait 10 secs - Checkpoint - wait 10 secs - Checkpoint - wait 10 secs...

vs.

Checkpoint - wait 10 secs - Checkpoint - wait 3 secs - Checkpoint/Lane change - ...

That would result in the lane change affecting the frequency at which torque must be seen, thus requiring more torque that otherwise might be enough for a nagless ride.

 

[diplomat33]

How many times do i have to be right?

Maybe we are talking in circles.

Let me explain this to you again. Periodically, the car checks to see if the driver has their hands on the wheel by checking for torque. Before every lane change, the car does an extra check to see if the driver still has their hands on the wheel by checking for torque. If it senses enough toque, then the driver passes the check and the car will then proceed with the auto lane change without any additional driver input. If the driver fails the check (not enough torque), then the car will prompt the driver to tug the wheel to confirm that they are holding the wheel before executing the auto lane change. So if the driver passes the check, the auto lane change happens with no additional driver confirmation. If the driver fails the check, then in that case, the car will require driver confirmation before executing the lane change. So the car does not require a confirmation for every single lane change as you claim, it only requires it if you fail the check.

That is what Musk is saying in that tweet and what multiple users have confirmed.

 

[electronblue]

Maybe we are talking in circles.

Let me explain this to you again. Periodically, the car checks to see if the driver has their hands on the wheel by checking for torque. Before every lane change, the car does an extra check to see if the driver still has their hands on the wheel by checking for torque. If it senses enough toque, then the driver passes the check and the car will then proceed with the auto lane change without any additional driver input. If the driver fails the check (not enough torque), then the car will prompt the driver to tug the wheel to confirm that they are holding the wheel before executing the auto lane change. So if the driver passes the check, the auto lane change happens with no additional driver confirmation. If the driver fails the check, then in that case, the car will require driver confirmation before executing the lane change. So the car does not require a confirmation for every single lane change as you claim, it only requires it if you fail the check.

That is what Musk is saying in that tweet and what multiple users have confirmed.

But the thing is — and Musk’s words do reopen this question — does the additional check cause it to require more torque than it used to?

If it is just all the same then it is all the same (same amount of torque applied overall during the drive suffices). If it is not the same and more action is in fact required than prior then it would not necessarily be quite as no confirmation ULC in the expected sense.

 

[rnortman]

Just to expound on your insightful post, I do understand the different ownership experience between these two groups. Basically, Group 1, like me, got a great car and got feature complete EAP pretty early in the life of their car. And if Tesla does achieve feature complete FSD with AP3, then Group 1 will also get feature complete FSD pretty early in the life of their car as well. Heck, new owners who buy a Tesla later this year, could conceivably drive off the delivery center parking lot with FSD already enabled. So yes, Group 1 will be very happy if Tesla achieves feature complete FSD later this year. For Group 2, however, it's a different story. They got a nice Tesla sure but they waited years for features that either never came or came late. They might have paid for FSD and got nothing for it yet. Even if Tesla does achieve feature complete FSD later this year with AP3, it will be much later in the life of their car. And even if they get the AP3 upgrade (which I believe they will if they are AP2), it will still come much later in the life of their car after a lot of missed promises. And AP1 owners are sadly out of luck at this point, I am afraid. So yeah, I can see why they would be less happy than Group 1.

You are missing Group X, of which I am a member. Group X always knew that Tesla's promises about truly driverless, robotaxi-level self-driving were complete and utter nonsense. Group X has personally experienced the tremendous strides made in L2 driver assistance features, such that after a "mere" 2 years AP2 caught up to roughly were AP1 was in 2016 and has recently begun to surpass it. Group X is pleased about this, but still angry at Tesla and Mr. Musk for continuing to outright lie about both timelines and future possibilities for L5 autonomy, because Group X recognizes that incremental software improvements in L2 ADAS features will not get you to safe and reliable L5 without corresponding changes in hardware. Group X is angry at Musk and his defenders for their cavalier and frankly ignorant attitude toward the development and release of safety-critical vehicle systems. Group X is angry at Tesla for pushing the envelope on safety in order to sell more cars and juice the stock and angry at his fanboys for lapping it up and promoting it all with irresponsible YouTube videos showing the car "driving itself". Group X thinks Mr. Musk has blood on his hands and should own up to that, but of course never will. Group X also thinks Musk and Tesla shareholders have profited enormously from his lies at the expense of Tesla's customers and that nobody will ever care to make those customers whole.

Group X is basically just mad at P.T. Barnum for being absolutely, completely right.

 

[diplomat33]

But the thing is — and Musk’s words do reopen this question — does the additional check cause it to require more torque than it used to?

If it is just all the same then it is all the same (same amount of torque applied overall during the drive suffices). If it is not the same and more action is in fact required than prior then it would not necessarily be quite as no confirmation ULC in the expected sense.

Good point. In any case, as Musk's tweet would seem to indicate, Tesla will probably remove this extra hands on wheel check at some point when they get more driver data that the auto lane changes are safe enough. The march of 9's as they say.

It is obvious to me that Tesla is being very cautious as they should be since they are taking an incremental approach towards further autonomy. They have to make sure each step is safe.

The stages of NOA seem to be:
1) stalk confirmation when the system was not good enough to do lane changes. This gave Tesla driver data to train the system for better auto lane changes.
2) No stalk confirmation once the system was good enough to do lane changes but an extra torque test was added just to make sure that the driver has their hands on wheel (to be safe just in case). In other words, the system does not need driver confirmation anymore but Tesla still wants the driver to keep their hands on the wheel just in case. Tesla can now collect even more data from auto lane changes that the driver is not confirming anymore to further validate how good the system is.
3) With even more data, when Tesla deems the auto lane changes safe enough not to require hands on wheel, then they will remove the extra torque test.
4) Eventually, I imagine Tesla will remove all torque tests for highway driving and then we will have hands free highway driving.

 

[Az_Rael]

I am curious what basic features of driver aids you consider Tesla is still struggling with?

TACC for one. We still own an AP 1 car so I get to compare regularly. TACC is amazing on AP 1 and buttery smooth compared to AP2.5 today. No random goosing the brakes and nice smooth slow downs in traffic. It is really too bad they split with Mobile Eye.

 

[rnortman]

By the way, we got another Musk tweet for "Group 1" to be excited about and for "Group 2" to scoff at:

Seems Musk is promising that Autopilot will eventually be able to handle potholes.

For the record, I think doing this with a reasonable level of accuracy is within the capabilities of their hardware, especially HW3, maybe even HW2 with some work if they prioritize this feature. Which they won't, not for a long time. And "reasonable level of accuracy" is a pretty big caveat -- you have to cope with the fact that because of false positives and the potential for serious consequences from false positives, it will never make any sudden maneuvers to avoid a pothole. It will not be willing to put its tires on the lane line to avoid a pothole. If you are following a car so that you only see the pothole at the last minute, the car will not have time to react safely and will just hit the pothole, when a human driver would have reacted.

Mostly I think that Musk choose to respond to this tweet, out of thousands of tweets he could have responded to, and chose to respond at this time in particular, when people have been asking for pothole avoidance for years, because he currently on a campaign to boost sales and the stock price by hyping FSD... again. But they have no intention of handling potholes any time soon.

 

[electronblue]

Group X is basically just mad at P.T. Barnum for being absolutely, completely right.

P.T. Barnum is @Bladerskb?

 

[electronblue]

In all seriousness though regarding @rnortman ’s ”Group X” it would be intellectually dishonest of me to claim I believed in the Level 5 capable hardware announcement by Tesla in 2016 either.

But I did believe in the FSD option buying me access to more active cameras and features soon (than EAP) and EAP coming out around December 2016 as well. There I was deceived by Tesla.

 

[mongo]

Well guess it depends.

Say the torque systems looks for that torque every 10 seconds..

Checkpoint - wait 10 secs - Checkpoint - wait 10 secs - Checkpoint - wait 10 secs...

vs.

Checkpoint - wait 10 secs - Checkpoint - wait 3 secs - Checkpoint/Lane change - ...

That would result in the lane change affecting the frequency at which torque must be seen, thus requiring more torque that otherwise might be enough for a nagless ride.

But that is not more torque, it is just more internal torque checks/ higher requirements.
If one always drove with one hand on the side of the wheel, then NoA has zero impact on their driving style (no additional input required from you). If one drove hands off and only taped a button when nagged, then the new system does increase the number of button pushes in toto (assuming land changes are not synchronous to nag period).
Only one of these two scenarios falls within the operational guidelines of Tesla's system...