So, when is APv2 going to be noticeably better than APv1? Or is it already?

[HankLloydRight]

Eh, while there's circumstantial evidence of this, where have they "admitted" it?

Ok, it's not an admission, but this is the biggest CYA I've seen in a long time.. I mean, it's HUUUUGE!

“However, we still expect to achieve full self-driving capability with safety more than twice as good as the average human driver without making any hardware changes to HW 2.0. If this does not turn out to be the case, which we think is highly unlikely, we will upgrade customers to the 2.5 computer at no cost.”

From: Tesla has a new Autopilot ‘2.5’ hardware suite with more computing power for autonomous driving

And yes, while this is from Seeking Alpha (with all due disclaimers), it does raise some good points about the NTHSA report w/r/t the AP1.0 fatal crash (the headline is a bit click-baity):

Did NTSB Decision Effectively Kill Off Tesla Autopilot Versions Prior To 2.5? - Tesla Motors (NASDAQ:TSLA) | Seeking Alpha

Namely that just using the steering wheel for driver alert detection is not sufficient... yet that's all that AP1.0 and 2.0 has (for now). Some other methods are going to be necessary.

Disclaimer: I'm long in TSLA (if it wasn't obvious).

 

[xyeahtony]

I think its ironic people pointing out AP2.0 is inferior to AP1.0 yet the only fatality ever recorded with Autopilot was on a 1.0 system and not 2.0. #shrug

 

[HankLloydRight]

I think its ironic people pointing out AP2.0 is inferior to AP1.0 yet the only fatality ever recorded with Autopilot was on a 1.0 system and not 2.0. #shrug

First, the flaws in the AP system as documented in the NTHSA investigation are essentially the same, or better in AP1.0. So keeping all things the same, the outcome of the accident would probably not be much different with AP2.0. As mentioned above, the problem was how Tesla detects driver attention (or lack thereof) which is essentially the same in both AP1 and AP2.

Second, correlation is not causation. Correlation does not imply causation - Wikipedia

 

[xyeahtony]

First, the flaws in the AP system as documented in the NTHSA investigation are essentially the same, or better in AP1.0. So keeping all things the same, the outcome of the accident would probably not be much different with AP2.0. As mentioned above, the problem was how Tesla detects driver attention (or lack thereof) which is essentially the same in both AP1 and AP2.

Second, correlation is not causation. Correlation does not imply causation - Wikipedia

oh believe me i was not implying AP1.0 was to blame any, just an interesting ironic twist. If anything i would just point out if AP2.0 was as horrible as some people like to make it (hyperbole is strong in the TMC) then i feel like the death count would be higher.

 

[verygreen]

oh believe me i was not implying AP1.0 was to blame any, just an interesting ironic twist. If anything i would just point out if AP2.0 was as horrible as some people like to make it (hyperbole is strong in the TMC) then i feel like the death count would be higher.

It's the other way around.
AP1 was so good people trusted in it more than they should have.
AP2 is so bad that people don't trust it at all and as such remain very attentive of the environment. In this sense AP2 ensures better driver engagement

 

[HankLloydRight]

oh believe me i was not implying AP1.0 was to blame any, just an interesting ironic twist. If anything i would just point out if AP2.0 was as horrible as some people like to make it (hyperbole is strong in the TMC) then i feel like the death count would be higher.

Ok, gotcha. Thanks for the explanation.

 

[No2DinosaurFuel]

It's the other way around.
AP1 was so good people trusted in it more than they should have.
AP2 is so bad that people don't trust it at all and as such remain very attentive of the environment. In this sense AP2 ensures better driver engagement

I think this is what tesla sorta did to AP1 also to ensure people are more engaged. I think I posted something about this a while back.

 

[oktane]

It's the other way around.
AP1 was so good people trusted in it more than they should have.
AP2 is so bad that people don't trust it at all and as such remain very attentive of the environment. In this sense AP2 ensures better driver engagement

Agreed. Hard for there to be a fatality when no one uses the feature. Total logical fallacy.

 

[NerdUno]

Tesla is already doing that with AP2.0 which they admitted can't support FSD, which they spent all of 2016 saying it absolutely would. Now all of a sudden it can't... and they're on the hook up upgrade tens-of-thousands of AP20 cars for free.

Wouldn't hold my breath on "free" hardware. Tesla will run the numbers and decide it is much more cost effective to refund the FSD deposits... and never mind the misrepresentations. They were only kidding.

 

[HankLloydRight]

Wouldn't hold my breath on "free" hardware. Tesla will run the numbers and decide it is much more cost effective to refund the FSD deposits... and never mind the misrepresentations. They were only kidding.

I agree, but I didn't want to be the first one to say it.

But in Tesla style, they won't refund the $ spend on FSD.. they'll instead offer you 1x or 2x of the $FSD off the price of a new car with AP(whatever).0

 

[oktane]

I agree, but I didn't want to be the first one to say it.

But in Tesla style, they won't refund the $ spend on FSD.. they'll instead offer you 1x or 2x of the $FSD off the price of a new car with AP(whatever).0

Exactly. A very calculated decision to defraud "rich" MS/MX customers out of money to give themselves working capital for M3 production. They will deal with the fallout later when M3 is generating cash. They will likely get away with just returning the FSD money, but I hope they have to pay penalties also.

 

[xyeahtony]

Agreed. Hard for there to be a fatality when no one uses the feature. Total logical fallacy.

except a lot of people do use the feature #shrug

 

[NerdUno]

I agree, but I didn't want to be the first one to say it.

But in Tesla style, they won't refund the $ spend on FSD.. they'll instead offer you 1x or 2x of the $FSD off the price of a new car with AP(whatever).0

Fool me once...

 

[ronnussbaum]

The auto pilot 2.0 sold with full self driving and enhanced auto pilot assures a guaranteed assurance of FULL SELF DRIVING. Now of course some less knowledgeable owners did not understand that there are federal hoops to jump through before the full feature set can be opened for use. Patience please for world changing technologies, I beg you have a little faith. Elon Musk has stated that a full self drive trip California to New York is in the offing. The time table for success is whatever it needs to be for this to happen and safely.

 

[brianman]

I'm certainly willing to concede that they haven't outright admitted it yet, but they did change course mid-stream to update the AP suite to add more cameras and sensors, which was my original point. If AP2.0 is fully capable of FSD, what are the extra cameras and sensors for?

Thought experiment:
1. AP and/or FSD team: Elon that there is a new hardware package that will improve reliability of the system by > 0% and is only marginally more expensive than the old package.
2. Elon: Does this mean the old package will not support EAP or FSD properly/fully?
3. AP and/or FSD team: No.
4. Elon: Ok, then don't put in the new hardware.

Really?

Improvements in hardware do not imply failings of old hardware. They imply opportunity and the general march of technology forward.

Your concern of the viability of AP2 hardware to support FSD might be right or at least well-founded, but the presence of newer hardware isn't evidence of it.

A silly example: They replaced the 40/80 charging with 48/72 charging. Do you consider that an upgrade, or that it implies that that 80a charging is broken? As an 80a owner that upgraded to a 72a capable vehicle it's definitely a downgrade, and my 80a Sig charging works just fine.

 

[Phenom300]

Folks there is a federal class action lawsuit over the lies and deception of autopilot, the video Tesla has on website was fabricated with about 80 edits, the car would kill me about 5 times each drive I make using autopilot if I do not override it. I plan to file my own lawsuit demanding full refund of car purchase of 115,000 or new car that actually works.
Tesla Owners Filed a Lawsuit Saying the New Autopilot Is 'Demonstrably Dangerous'

 

[iaw4]

I am not happy with AP2, either. I purchased my S [and ordered a 3] because I care about AP. With better hardware and software upgrades, I have hope.

I only wish (1) the AP could hold the lane on Sunset drive reliably [which, given how many Teslas are driving within lane on sunset every day, should be easy], and (2) deal well with cars on neighboring lanes that cross into my lane. this is not good even on the freeway, where it should be easy. (at times, the car also has tended to want to lurk suddenly.) .

alas, let's give them some time. it's hard to keep self-driving engineering teams intact at the momen. we hope it will improve.

if there is one aspect worth complaining about, it is the lack of communication from tesla with its owners (other than sales claims, of course). I have no idea of what the current AP version is, nor when they plan to release the next major and minor versions. sort of completely in the dark...

 

[martinwinlow]

I had a loaner Model X last weekend and in the hour's play I had on the M25 (motorway) it's Ap not as good as my 2015 Model S' Ap. The latter is much smoother.

 

[Sunshine State]

I think its ironic people pointing out AP2.0 is inferior to AP1.0 yet the only fatality ever recorded with Autopilot was on a 1.0 system and not 2.0. #shrug

There are more AP1 cars and those cars have logged more miles so just by that standard give the AP2 more time and it is inevitable that the accidents will increase. It's not a question of which works better, AP1 is clearly superior in performance. I wanted to purchase an AP2 S after enjoying the AP1 in our new model X P90D for a few months and expected Tesla to work out the problems. After weekend demo driving a new AP2 Model S we realized the benefits of the AP1 system and purchased a second Tesla, a loaded 2016 AP1 Model S instead. AP2 is not smooth on the highway compared to AP1, AP2 randomly dives for exits and turns gentle bends in the road into 20-30 short straightaways with constant small corrections rather than one smooth radius. I can understand a few owners thinking AP2 is fine but until you have lived with a better system you just don't know different, it's just a shame Tesla was forced to take a step or ten backwards.

 

[outie]

except a lot of people do use the feature #shrug

They do, with even more attention on AP behavior than driving without AP because it’s constantly trying to kill them and they know it. If they let AP do its own things the fatality rate would skyrocket.