Ars: Tesla has a self-driving strategy other companies abandoned years ago

[john85775]

Pretty damning article, essentially taking the continuous Tesla walk-back on FSD and AutoPilot language and putting it in the context of the other self-driving approaches. It's worth a read regardless of your stance on Waymo, LIDAR, Uber et. al.

Tesla has a self-driving strategy other companies abandoned years ago

The new pricing structure defines full self-driving differently. The ability to navigate freeway interchanges, for example, was shifted from "Enhanced Autopilot" in the old pricing structure to "Full Self-Driving" in the new one. Later this year, Teslas with the "Full Self-Driving" package will be able to "recognize and respond to traffic lights and stop signs" and perform "automatic driving on city streets."

Hence, Tesla now seems to define "full self-driving" as a system that can handle most road conditions under the supervision of a human driver. Tesla is still aiming to improve the system enough that—eventually—it can operate without human supervision. But the new pricing structure makes things less awkward in the meantime, since Tesla can now argue that customers have already received "full self-driving" features like the ability to stop at stop signs.

 

[Tam]

...Pretty damning article...

There's nothing new about anti-Tesla view as cited by the article.

Most companies like Waymo believe that human is not to be trusted to drive. It believes that it is much worse if you give human an incomplete automation system because they may fall asleep and die.

There's another view who thinks it is immoral to deprive the public of technology despite no matter how incomplete it is.

If we listen the first group like Waymo, then we wouldn't have commercial aircraft autopilot because that system still results in crashes and imperfect too. And of course, there were times pilots were caught sleeping while on autopilot too!

But if you are pro-Waymo way, then don't drive and wait till Waymo perfects its technology. That's your choice, there's nothing wrong with choosing what you prefer.

About the naming of Full Self Driving: That's the feature owners can pre-pay before the reality will arrive.

When people pay, whether the technology is here or not, Full Self Driving is what they pay for.

It is true that Full Self Driving technology is just not here yet, so it doesn't matter how much people pay, if it's not here, it's no here.

Pay now and wait for technology will mature and get what you paid for.

One good thing about Tesla is: Even when the technology is not here yet, you can still grow with the system, little by little until you'll get what you paid for.

 

[john85775]

Tesla’s approach relies on lane markings. EAP has put me dangerously close in freeway driving (carpool lane) to hitting concrete pylons that jutt out into the lane - so the story of the driver that crashed when the lane markings were off resonated.

Lane splits are another - when one lane becomes two, the car seeks center (whilst one is becoming two) then abruptly corrects.

Despite all this “learning”, those two dangerous instances (along with no-explanation slow downs or a lack of awareness of an impending merging car) make what Tesla NOW calls “full self driving “ a bit of a joke.

History will judge, but Elon doubling down on his approach without flexibility may prove to be problematic.

 

[Tam]

...“full self driving “ a bit of a joke...

What Tesla is selling is the future.

In future, its car won't hit the concrete, but the technology is not here just yet so that is why there's a requirement of an attentive driver.

If you can't fulfill that requirement, then don't use an incomplete product.

Why selling FSD when it is not FSD?

It's just like asking why selling a ticket to Mars and not calling it a ticket to earth?

Yes, it's possible to be on earth now but not possible to be on Mars!

If I paid for a ticket for Mars, I don't want that to be renamed as "Earth" because that's not what I paid for.

 

[more863-also]

The fact that Tesla considers NoA a feature worth moving to FSD is comical based on how bad it is. If that's the path to FSD, Level 3 autonomy will never happen without major changes.

 

[ItsNotAboutTheMoney]

Most companies like Waymo believe that human is not to be trusted to drive. It believes that it is much worse if you give human an incomplete automation system because they may fall asleep and die.

Waymo believes that as an advertising company it won't make as much money from driver assistance.

 

[ItsNotAboutTheMoney]

The fact that Tesla considers NoA a feature worth moving to FSD is comical based on how bad it is. If that's the path to FSD, Level 3 autonomy will never happen without major changes.

Tesla believes that it wants to sell Autopilot for $3k instead of EAP for $5k, because of the shift in sales emphasis to entry-level premium, so it went back to Autopilot being adaptive cruise and autosteer.

 

[erik_k]

Lane splits are another - when one lane becomes two, the car seeks center (whilst one is becoming two) then abruptly corrects.

Our 2017 Honda Accord Hybrid does the exact same thing. At least Tesla is pushing updates OTA so things are still getting better with AP2+.
While the tech is improving I’m noticing a lot of exits where they’re now painting white exit stripes when the exit lane splits. This keeps the right lane uniform with markings on the left and right side of the lane. This is helping a lot, but driver hands on the wheel and eyes up are still going to be necessary for quite a while.

 

[john85775]

What Tesla is selling is the future.

In future, its car won't hit the concrete, but the technology is not here just yet so that is why there's a requirement of an attentive driver.

If you can't fulfill that requirement, then don't use an incomplete product.

Why selling FSD when it is not FSD?

It's just like asking why selling a ticket to Mars and not calling it a ticket to earth?

Yes, it's possible to be on earth now but not possible to be on Mars!

If I paid for a ticket for Mars, I don't want that to be renamed as "Earth" because that's not what I paid for.

Don’t pee on my leg and tell me it’s raining. I was all in on this vision until parallel parking and summon (ok so it goes back and forth when I press the buttons) were lumped in with “Full self driving.”

 

[more863-also]

Our 2017 Honda Accord Hybrid does the exact same thing. At least Tesla is pushing updates OTA so things are still getting better with AP2+.
While the tech is improving I’m noticing a lot of exits where they’re now painting white exit stripes when the exit lane splits. This keeps the right lane uniform with markings on the left and right side of the lane. This is helping a lot, but driver hands on the wheel and eyes up are still going to be necessary for quite a while.

Yeah but your Honda cost way less and probably still hasn't dropped $4-8k in value. Fact is that AP isn't really much better than Honda Sensing or Nissan ProPilot.

 

[bhzmark]

Yeah but your Honda cost way less and probably still hasn't dropped $4-8k in value. Fact is that AP isn't really much better than Honda Sensing or Nissan ProPilot.

Consumer Reports ranked Tesla Autopilot tops in capability and performance. It only preferred the Cadilac because it has more nanny limitations. Cadillac Tops Tesla in Consumer Reports' First Ranking of Automated Driving Systems

The new Tesla AP technology killed BMW MB and Volvo, and even the older AP tech in the Model S that IIHS tested did better except on hills.
Tests uncover issues for advanced features

And Honda did poorly compared to others also: Comparing lane departure prevention systems

 

[more863-also]

Yeah that's not been my experience. ProPilot is pretty damn good. Plus the CR rankings are nonsense and ding you for insufficient nags and "ease of use" which is very subjective, it's not hard to turn on PP. Oh and the entire tech package on a Leaf costs what AP does. Interesting that IIHS testing shows the Model S to be the most dangerous of all, crossing lines frequently.

As a former 3 owner my vehicle absolutely touched the line on a daily basis and crossed it a few times too, showing 0 fails simply isn't what I experienced. I had at least 3-4 critical disengagements (red hands) per month, are you really saying AP never failed for you?

 

[bhzmark]

Yeah that's not been my experience. ProPilot is pretty damn good. Plus the CR rankings are nonsense and ding you for insufficient nags and "ease of use" which is very subjective, it's not hard to turn on PP. Oh and the entire tech package on a Leaf costs what AP does. Interesting that IIHS testing shows the Model S to be the most dangerous of all, crossing lines frequently.

As a former 3 owner my vehicle absolutely touched the line on a daily basis and crossed it a few times too, showing 0 fails simply isn't what I experienced. I had at least 3-4 critical disengagements (red hands) per month, are you really saying AP never failed for you?

The CR RAnkings just are what they are. The capability and performance was a test of actual lane-keeping. Tesla beat everyone on that. The nags and eye camera and user interface were other categories that I agree are much less important, yet CR seems to have weighted them enough to let the worse performing Supercruise rank higher overall than the better performing Tesla Autopilot.

The other tests just are what they are -- certainly more credible and reliable than someone on the internet.

And no I'M not saying anything about my own experience with AP -- I'm simply sharing factual information from independent organizations that are in the business of testing these things and their results are credible and show the superiority of Tesla AP lane-keeping ability.

 

[john85775]

It only preferred the Cadilac because it has more nanny limitations.

Those "nanny limitations" are there, literally, to save your life. Tesla's AP largely depends on lane markings, and as described in several instances above, can fail and do so abruptly. If you aren't supervising AP, your risk level goes WAY up because it simply isn't mature nor capable enough to use without supervision. If you want to gamble your own life on it, so be it - but please don't put the rest of us at risk.

 

[more863-also]

Lol, anyone who has extensively used EAP knows it frequently touches lines and sometimes crosses them in turns. According to that "factual" information that should never happen. And none of that IIHS information covers ProPilot.

 

[john85775]

Here are the gaps I experience on a regular basis with EAP on our November 2018 build Model X:

- One lane becoming two - what should I do?
- #1/carpool lane - "hairpin turn of death" (inches from the concrete pylons)
- Adjacent vehicle merging ahead of you (... and... here it comes.... BRAKE SLAM!)
- Random slow down - just because
- Lane-change-just-kidding-half-way-through-lets-go-back-to-the-old-lane
- BEEP BEEP BEEP YOU'RE GOING TO CRASH (never mind, that's just a parked car on a curvy road)

Don't even get me started with Navigate on Autopilot.

Do I use AP? Yes. Would I trust my life with it? HELL NO.

 

[Tam]

...Would I trust my life with it? HELL NO.

That is very healthy Tesla Owner's Manual Compliant.

The manual keeps repeating the mantra "Do not rely" over and over again in the hope that someone can comply.

 

[Tam]

...FSD is comical based on how bad it is...

You are buying an unrealized future that has not been fulfilled.

When I bought mine, Autopilot was not functional in city streets. As a matter of fact, it could operate on freeways with a speed of 35MPH or under while the highways around here are posted 70 MPH.

Comical?

If you don't like heat then don't become a firefighter and complain the flames are hot!

If you don't want an unfinished product, then don't pay for the beta product!

 

[bhzmark]

To the extent assisted driving becomes overall statistically safer than human driving, the limitation to only mapped out roads will cost rather than save lives. Cadillacs decision will result in more accidents outside the geofence without the benefit of the assistance. Although maybe for Cadillac it just isn't good enough to be safer outside the geo fence.

Those "nanny limitations" are there, literally, to save your life. Tesla's AP largely depends on lane markings, and as described in several instances above, can fail and do so abruptly. If you aren't supervising AP, your risk level goes WAY up because it simply isn't mature nor capable enough to use without supervision. If you want to gamble your own life on it, so be it - but please don't put the rest of us at risk.

 

[neroden]

Pretty damning article, essentially taking the continuous Tesla walk-back on FSD and AutoPilot language and putting it in the context of the other self-driving approaches. It's worth a read regardless of your stance on Waymo, LIDAR, Uber et. al.

Tesla has a self-driving strategy other companies abandoned years ago

FWIW, I more or less agree with the article, but I'd add that the other companies don't have a chance in hell of getting full-self-driving either *and* are already killing people.

I'm not sure there's a real way around the attention problem. I think the correct implementation is effectively "cars which refuse to crash into anything", and leave it at that -- the driver has to do all the driving, but if the driver tries to crash into something / run a red light / etc., the car will not accelerate, it'll stop. This is basically how safety systems got implemented on trains.