"Temporarily" Better Be A Short Time, Or Tesla Will Again Be Over-promising And Under-Delivering

[Andyw2100]

I think today's announcement about what is effectively AP 2.0 is fantastic. It sounds like Tesla will be delivering much more than most people expected in the very near future.

I do have a concern, though, about the following, from the blog post:

"Before activating the features enabled by the new hardware, we will further calibrate the system using millions of miles of real-world driving to ensure significant improvements to safety and convenience. While this is occurring, Teslas with new hardware will temporarily lack certain features currently available on Teslas with first-generation Autopilot hardware, including some standard safety features such as automatic emergency braking, collision warning, lane holding and active cruise control. As these features are robustly validated we will enable them over the air, together with a rapidly expanding set of entirely new features."

Those features that will be "temporarily lacking" are features that the new owners-to-be paid for, and counted on having. Yes, they may be getting more than they bargained for eventually, getting the option to utilize AP 2.0 instead of the current Autopilot features. If "temporarily" is a short time, I expect everything will, for the most part, be fine. But if things drag on, as some promised and paid for features have in the past, I imagine some of these new owners will grow impatient about not having features that they ordered and paid for.

Let's just hope "temporarily" is a short time.

 

[JRD1]

The articles I've read say Elon said December in the Q&A section of the call. Haven't seen a transcript to know for sure though.

 

[S4WRXTTCS]

In the short term it means most of the arguments on TMC will go like this

Person A: It's going to be at least a year before AP 2.0 reaches feature parity/performance as AP 1.0
Person B: No, the website says Dec 2016
Person A: You're smoking crack if you believe that
Person B: You don't have to be such an ass.

Where the difference between person A and person B is person A experienced the year or so wait for AP to get anywhere close to promised.

As to the rollout of features to AP 2.0 I think it's going to be gradual with it overtaking AP 1.0 in features/performance within a year. The first release or two will likely have a lot of glitches that I'd rather not be a part of.

TACC which I care most about is mostly reliant on radar so I'm not expecting much difference with that one.

What I'll be the most jealous of: Side Monitoring that works
What I won't be jealous of: False positive AEB activation.

 

[S'toon]

For a time, cars with the new hardware will have less capability to assist drivers with steering or braking than older models running Tesla's Autopilot technology, Musk said. By December, he said, he expects the newer models to reach parity with the older vehicles.

All new Tesla vehicles will have self-driving technology, Elon Musk says

 

[Andyw2100]

All new Tesla vehicles will have self-driving technology, Elon Musk says

From the article you referenced:

"For a time, cars with the new hardware will have less capability to assist drivers with steering or braking than older models running Tesla's Autopilot technology, Musk said. By December, he said, he expects the newer models to reach parity with the older vehicles."

So Musk is suggesting December as the time when the new cars should reach parity with the older cars. I hope Tesla meets that date.

 

[JRD1]

It's also in the WSJ article Elon just linked to on Twitter

 

[Spidy]

"Temporarily" Better Be A Short Time, Or Tesla Will Again Be Over-promising And Under-Delivering
That was my first thought when I read that part, too. Remember?
Autopilot lane keeping still not available over 6 months after delivery

 

[bob_p]

Tesla operates on a very rapid development cycle - it's likely they've done enough testing of the new sensors to verify that the combination of sensors and processor are sufficient to address any shortfalls they found with AP 1.0 - and they wanted to get this hardware out to the market as quickly as possible, for several reasons:

1. Tesla is getting negative publicity on accidents with AP 1.0. While they are making improvements in 8.0 and 8.1, there are limitations due to the sensors are lack of processing power. By deploying 2.0 now, if there are any additional accidents with 1.0, they can claim "the new cars being sold now have improved AP and will fix these problems" (even if it may take several years for that to be completely true).
   
   
2. The market expected AP 2.0 to come out soon. If they delayed the hardware rollout too much, they risked impacting short term sales. By doing this at the beginning of the quarter, they will have some customers with pending orders disrupted, but for most of the quarter, they'll be selling AP 2.0 cars.
   
   
3. And, the big reason to get the hardware out as soon as possible - Tesla will begin collecting data from an increasing number of cars, almost immediately - and that data will help them test and improve the AP software.

Assuming Tesla only finalized the hardware combination a few months ago (which is very likely), it's understandable that they haven't completed testing of all of the AP 1.0 features. They likely verified the sensors are working at least as well as they expected and they're confident the hardware is fine. But it will take time to adjust the software for the new inputs - and then to fully test the software changes with the new hardware, before enabling the complete AP 1.0 feature set.

And rather than delay the hardware introduction (for the reasons above), they've put the new hardware into production - and by the time most people get the AP 2.0 cars they are ordering now, it's likely that most, if not all, of the AP 1.0 features will be working.

 

[Gwgan]

There is something that feels very wrong about the timing. Maybe they wanted to get out in front of people noticing the new hardware, maybe they wanted to make a big deal about validation, but to announce the upgrade and lose safety features (temporarily) seems like the did it now because they had to. Why might they have had to? Mobileye won’t let them have any more chips? NTSB thought it would be a "really good idea"? I would expect instead that T would announce the L5 future possibility, declare all new cars have the hardware, and tell recent buyers and those in the queue that they could have the Enhanced AP software for a few additional thousands but that current safety features would be included but even better.
It also seems presumptive for them to be setting a price, taking money, and locking in the hardware set for L5 when things could change so much between now and then. Remember when the X was supposed to be able to be summoned around corners and take highway ramps but all it could do before AP2.0 was back up a couple car lengths in a straight line?
I am so glad my M3 will have L5AP so I can pimp it out and pay for my “driver’s car MX” and I fully expect the sensors, and maybe even the chip, to improve between today’s AP2.0 and late 2017.

 

[Andyw2100]

And rather than delay the hardware introduction (for the reasons above), they've put the new hardware into production - and by the time most people get the AP 2.0 cars they are ordering now, it's likely that most, if not all, of the AP 1.0 features will be working.

The problem is there are people who ordered their cars two, three, or perhaps four weeks ago (I'm not up on the production timeline right now) and those cars are just going into production now. Those cars will be delivered in another few weeks. So some people who placed their orders in September or early October, and who may be taking delivery in early to mid November will be taking delivery of cars that have, at least for the time being, less capability than what they ordered and paid for. Sure, down the road--hopefully a short road--the cars will actually be capable of more than what these customers paid for. But at the time they take delivery the product will not have the functionality promised.

I see this as an indication that the mentality at Tesla that thinks this is an OK way to operate has not changed. I had hoped Tesla would learn from its mistakes, and try to make sure not to repeat them. I recognize that this was a tough situation, and Tesla couldn't really call all the affected customers and ask them which they would prefer--AP 1, delivered as ordered, or AP 2, better hardware, but temporarily less functionality. Doing so would have required building cars with both sets of hardware for a while, and clearly that's not efficient, and maybe not even possible. I don't know what the better solution in this case could have been, but I have to think that there was a better option than delivering cars to customers that don't do what Tesla said they would do at the time the customers ordered them. The fact that Tesla continues to think this is OK, even after all that has taken place, concerns me.

Perhaps my greatest concern is what will happen if Tesla maintains this philosophy when it starts to deliver the Model 3. Model 3 is supposed to be for the mainstream market. The mainstream market will not accept less than what they paid for even on a "temporary" basis.

 

[Diehard7]

My question that I may post in another thread is will A1.0 be able to learn from the advanced suite installed in cars w/A.2.0? My wet dream is that considering the roads are the same & the cars are traveling many of the same roads will Tesla share the fleet learning between the 2 versions? It would be dope if the A1.0 slow kids could cheat off the test of the A2.0 kids to better negotiate turns and roads.

 

[bob_p]

Disabling of functionality has happened before with Tesla.

We had a period when the air suspension systems were disabled after a car drove over a trailer hitch on the road and had a fire. Tesla immediately disabled the air suspension systems and then turned them back on after they completed their investigation and did a retrofit adding more armor plating underneath the battery packs.

Yes, at that time, those of us who had the air suspension systems were frustrated in that we had paid for the feature - and it had been turned off using a OTA update - that was deployed so quickly that we didn't realize the feature was gone until after almost everyone had the update installed.

For those people with pending orders and getting AP 2.0 after they ordered AP 1.0 - yes, they may have a few months when AP 2.0 is missing a few features from AP 1.0.

But that will likely only be a few months, until Tesla completes testing of AP 2.0 with the AP 1.0 functionality.

And then after that, this group of owners will have the benefits of AP 2.0 and the potential to see self driving in their cars, something that would not be possible if they received cars with AP 1.0.

The alternative would have been for Tesla to delay AP 2.0 introduction for another 3 months - and introduce it after they had AP 2.0 tested. And, if they had done that, this group of owners would have received fully functioning AP 1.0 hardware - but they would have been disappointed to find out in a few months that they had just missed getting AP 2.0 on their cars.

Rather than being disappointed in missing functionality for a few months, the affected owners appear to be getting a pretty good deal - and in the long run, they should benefit from the early introduction of the hardware, before the software was fully operational.

 

[Vitold]

There is something that feels very wrong about the timing. Maybe they wanted to get out in front of people noticing the new hardware, maybe they wanted to make a big deal about validation, but to announce the upgrade and lose safety features (temporarily) seems like the did it now because they had to. Why might they have had to? Mobileye won’t let them have any more chips? NTSB thought it would be a "really good idea"? I would expect instead that T would announce the L5 future possibility, declare all new cars have the hardware, and tell recent buyers and those in the queue that they could have the Enhanced AP software for a few additional thousands but that current safety features would be included but even better.
It also seems presumptive for them to be setting a price, taking money, and locking in the hardware set for L5 when things could change so much between now and then. Remember when the X was supposed to be able to be summoned around corners and take highway ramps but all it could do before AP2.0 was back up a couple car lengths in a straight line?
I am so glad my M3 will have L5AP so I can pimp it out and pay for my “driver’s car MX” and I fully expect the sensors, and maybe even the chip, to improve between today’s AP2.0 and late 2017.

It's a speculation, but it's quite possible that Tesla had a two year agreement with Mobilieye that ended in October (AP1 was introduced in October of 2014). In any case I believe Tesla planned to end the relationship at this time, as it makes a lot of sense to do so now than later, and AP2.0 is not an ad-hoc implementation.

After 8.0 deployment Tesla may have decided to run AP2.0 in shadow mode for longer just to be certain everything is functioning better than in AP1.0. After-all they set pretty high bar for themselves with hardware 1.

I think a good indicator if above is true will be how soon TACC will be enabled with AP2.0.

 

[Boatguy]

The problem is there are people who ordered their cars two, three, or perhaps four weeks ago (I'm not up on the production timeline right now) and those cars are just going into production now. Those cars will be delivered in another few weeks. So some people who placed their orders in September or early October, and who may be taking delivery in early to mid November will be taking delivery of cars that have, at least for the time being, less capability than what they ordered and paid for. Sure, down the road--hopefully a short road--the cars will actually be capable of more than what these customers paid for. But at the time they take delivery the product will not have the functionality promised.

I see this as an indication that the mentality at Tesla that thinks this is an OK way to operate has not changed. I had hoped Tesla would learn from its mistakes, and try to make sure not to repeat them. I recognize that this was a tough situation, and Tesla couldn't really call all the affected customers and ask them which they would prefer--AP 1, delivered as ordered, or AP 2, better hardware, but temporarily less functionality. Doing so would have required building cars with both sets of hardware for a while, and clearly that's not efficient, and maybe not even possible. I don't know what the better solution in this case could have been, but I have to think that there was a better option than delivering cars to customers that don't do what Tesla said they would do at the time the customers ordered them. The fact that Tesla continues to think this is OK, even after all that has taken place, concerns me.

Perhaps my greatest concern is what will happen if Tesla maintains this philosophy when it starts to deliver the Model 3. Model 3 is supposed to be for the mainstream market. The mainstream market will not accept less than what they paid for even on a "temporary" basis.

If your car is somewhere between a confirmed order and delivered, it's essentially a crap shoot what h/w will show up in your car. I would guess there are a lot of people pounding on Tesla to get an answer to that question right now. Some are going to be happy, others not so much, as I wasn't when I took delivery 18hrs before they announced the refresh (which they hid during the factory tour). Tesla has learned nothing about managing customers.

The AP2.0 h/w announcement is all about selling more cars, now, and collecting the cash from customers, before Tesla can actually deliver the product it's selling because "it's just software". Besides the obvious problems that Tesla has making software on anything close to the schedules that Elon announces, there is an implicit assumption that the h/w world is static. The reality is that Nvidia will have a next generation of h/w before Tesla is delivers the majority of the functionality that Elon promised. Most M3's will come with a next generation of Nvidia h/w, not whatever they are placing in the MS/MX today.

Buy today and your h/w will be out of date before the s/w required to make it useful is delivered.

 

[Max*]

I wasn't when I took delivery 18hrs before they announced the refresh (which they hid during the factory tour). Tesla has learned nothing about managing customers.

So they should Osborne [effect] their product line?

 

[Boatguy]

So they should Osborne [effect] their product line?

They may have just done that by telling everyone taking delivery for the next few months that their car will not have the same degree of AP functionality as cars that were delivered earlier. But pay us $5,000 today and you may have a better AP in three months, IF we complete our s/w development on time (which we've never done) AND "subject to regulatory approval" (which means getting through a process that no manufacturer has traversed and governments have little experience with). Until that happens, your new Tesla will have less AP functionality than an AP 1.0 car.

The real issue will be with someone who ordered the car expecting to get AP 1.0 functionality and is now going to take delivery of a car without it, but a promise that it will be much better in the future. The Q4 numbers will tell us how the market feels about that.

With v8 their shipping navigation and media player that regressed, now with AP2.0 they are shipping AP which has regressed functionality. This is not progress.

 

[hiroshiy]

So:

No lane holding = no autopilot
No active cruise control = no TACC

for three months in the US? Hope disabling AEBs won't affect insurance rates.

 

[vgrinshpun]

Just for information, estimated delivery time shown on Tesla's ordering page for customers in US went from late October couple of weeks ago straight to December, skipping November. Estimated time for European deliveries is late January.

If Tesla stays on schedule with delivering AP 1.0 functionality for cars with AP 2.0 hardware in December, all owners of cars with AP 2.0 hardware will have it functional by the time they receive the cars.

 

[AWDtsla]

At this point Tesla has had enough product releases to say there is a very very consistent pattern. Features promised and paid for in one model, don't actually happen until the next model. The software will be delivered late, and will underwhelm. The thing you thought it would enable, won't actually work *right* until you buy a new car.

Note how autopilot is no longer autopilot, that's lane holding. Because AP 1.0 can't really deliver.

 

[Topher]

, but to announce the upgrade and lose safety features (temporarily) seems like the did it now because they had to.

They had to. They need a few tens of million of miles of actual driving to calibrate the new sensors to the old algorithms. They can't do that without a lot of cars with the new hardware, on actual roads. They could do it with a 100 test cars and a year of driving, or with 10,000 cars and a month or two. In the meantime 100,000 cars either get the old system, or a new system that doesn't work.

Thank you kindly.