FSD may require a hardware upgrade...

[schonelucht]

Throw in a slam dunk consumer fraud case and Tesla really better deliver free of charge or they will drown in litigation. Remember, I used to do these lawsuits as a young attorney because attorneys fees are baked in to most state consumer fraud laws. Nothing better than a slam dunk case where the other side fights (its like those Chinese finger traps -- the more you struggle the worse it becomes).

Yet nothing has come of that with earlier broken sales promises by Tesla. Single exception are the Norwegian P85 owners who got a consideration for missing the promised power. It seems Tesla's conditionals are strong enough to keep the lawsuits from materializing.

 

[Spidy]

Yet nothing has come of that with earlier broken sales promises by Tesla. Single exception are the Norwegian P85 owners who got a consideration for missing the promised power. It seems Tesla's conditionals are strong enough to keep the lawsuits from materializing.

According to some people in the German TFF forum a few owners also got compensated, but only once they lawyered up and then with a NDA.

 

[croman]

I remain bullish that Tesla will deliver at least on some aspects of EAP and FSD this year.

I think if they are able to use the current hardware as they indicated then there isn't much anyone can do about a very slow rollout.

 

[u00mem9]

At the risk of being pedantic - if you don't understand the fundamental difference between Mobileye and Nvidia it's easy to think Tesla is behind. Nvidia's system learns from scratch. Mobileye came with a pre-baked set of items it could recognize. Nevertheless the Mobileye-based AP 1 hardware released in Oct 2014 was not enabled for 12 MONTHS after the hardware shipped. Do you have an autopilot Tesla? If so - do you remember what 7.0 was like back in October 2015? It was very sketchy compared to what we have today - February 2017. It is pretty darn rock solid now - but it's been almost 2.5 years since AP 1.0's fleet learning first began and the cars were running shadow mode.

AP 2.0 hardware has been in the wild - running a brand new neural network that was essentially an infant and is designed by Nvidia to learn from scratch - for less than 4 months. 2.5 years vs 4 months. And yet it is iterating and progressing far faster than AP 1.0 did. Which it should - it's a much more powerful computer.

Your sig says you joined in June 2016 - which means you were not around this neighborhood to watch all the crying and moaning folks did while AP 1.0 was being developed. As soon as it was released - everybody went silent.

Tell me what other company will be ahead of Tesla in a marketed product 24 months from now? Who? Mercedes? Really? What massive fleet does Mercedes have on the roads finding corner cases for them now? Tesla is adding 25,000 cars a quarter to worldwide roads gathering data.

24 months from now Tesla will likely release Autopilot 3.0 - but by then even if the competition has released a fleet learning program - Tesla will have billions of miles accumulated on Nvidia's platform, all of which will be used to enhance the performance of the third generation.

I think Tesla has this game on lock - they're accelerating away from the competition.

I bought AP1.0 when it had reached a proven level of stability (plus a couple months for delivery). I will buy an autonomous vehicle as soon as they reach the same level. At this point, the signal:noise coming from Muskco indicates that probably won't be a Tesla. If you think it was ugly from Oct 2014 to Oct 2015...get ready for some next-level gnashing of teeth after another 6 months of empty promises and cryptic tweets.

Btw, your attempts to explain what you know about how these systems work reveal a little more about your depth than you would probably intend. Kinda 'Layman level slides for the salesman' knowledge. No offense...but I think if you had a bit more perspective, you would have a bit less to say.

 

[davidc18]

Love - Love - Love! AP 1.0 on FW 7.1 (The roads here in south florida are perfect for the system)
I was initially excited about AP 2.0. Now I am back on earth outside of the Tesla reality distortion field and I can see that AP 2.0 will be just like AP 1.0 and only deliver on part of the initial promises. No one will get anything from Tesla other than some broken promises and over time, everyone will move on when AP 3.0 is announced. Nothing new here. Enjoy the car you have today. Welcome to the Tesla club.

 

[calisnow]

I bought AP1.0 when it had reached a proven level of stability (plus a couple months for delivery). I will buy an autonomous vehicle as soon as they reach the same level. At this point, the signal:noise coming from Muskco indicates that probably won't be a Tesla. If you think it was ugly from Oct 2014 to Oct 2015...get ready for some next-level gnashing of teeth after another 6 months of empty promises and cryptic tweets.

Btw, your attempts to explain what you know about how these systems work reveal a little more about your depth than you would probably intend. Kinda 'Layman level slides for the salesman' knowledge. No offense...but I think if you had a bit more perspective, you would have a bit less to say.

Okay bud - here's less. You haven't answered the question - who will beat Tesla to market 24 months from now with full autonomy and how?

As for the ratio of wordiness to knowledge - whether or not you're right you haven't explained why I'm wrong. So far bud you're batting zero.

 

[bob_p]

Just received the e-mail that our new S 100D has completed production, so we'll get our EAP/FSD system in the next few weeks!

It is surprising Tesla's attorneys aren't being more careful about Tesla's exposure on FSD. If Tesla determines that hardware upgrades are needed to get FSD working, they will either have to provide free hardware upgrades or risk major fallout from the rapidly growing number of customers who believe they have purchased FSD with their cars.

Perhaps, they should consider the following:

1. Update the Tesla website and purchase terms to admit hardware upgrades may be required to get FSD operational.
2. For owners purchasing FSD, if hardware upgrades are required for FSD activation, Tesla will either provide those upgrades for free, or Tesla will provide a significant credit towards the purchase of a new Tesla vehicle that has operational FSD ($10K?).
3. For owners deciding to defer the purchase of FSD, the price and availability of the FSD activation option is not guaranteed. If a hardware upgrade is required for FSD activation, either the upgrade price could be increased, or if the upgrade is too expensive, the FSD option could be withdrawn.

 

[disagree]

I expect two things:

1. A hardware update will be needed for FSD -- and Tesla will try to accommodate.

2. A hardware update will be needed if Tesla wants to realize EAP -- and Tesla will have a big problem.

 

[Matt125]

I always thought this would the case. FSD on HW2, given the problems they've already been having with HW2, seemed a bit far-fetched.

I won't mind if the Model 3 is still on HW2 when I get mine, however. I'm unlikely to ever bother with FSD - I enjoy driving and wouldn't want the car to do it for me, and Autopilot will give me all the help I need for long trips/boring traffic.

For others, I would recommend they hold off buying FSD until it is tried and tested, and if you're expecting to get the full FSD experience on a Tesla you buy right now, you'll probably be disappointed.

 

[Sir Guacamolaf]

I expect two things:

1. A hardware update will be needed for FSD -- and Tesla will try to accommodate.

2. A hardware update will be needed if Tesla wants to realize EAP -- and Tesla will have a big problem.

I @disagree ;-)

Tesla has two very good reasons not to focus on upgrading your car,
- It diverts attention and resources from creating the next new toy. Building in upgradability is a cost.
- It serves as a disincentive to prospective buyers to buy a new car.

History has shown that the only way to upgrade Tesla, is to buy a new one. I don't expect this to change.

One other thing I'd like to add

I paid $2500 for AP1. Compared to what you have to pay now, vs. what you get - holy moly did I get a steal of a deal. AP1, no regrets. And I don't think AP2 will be significantly better anytime soon, even when it is, practically speaking it's not going to make a big difference as long as I still need to focus on the road as if my life depended on it.

 

[ChadFeldheimer]

If anyone thinks that Tesla would ever announce that AP2 HW is insufficient to implement FSD capability...

I'll bet you anything that I will (someday) flap my arms and fly away. Collection terms: you are welcome to collect on your bet when I choose to announce defeat (or declare bankruptcy... good luck with that!).

Where this analogy breaks down: I do believe Tesla will, in good faith, attempt to deliver FSD functionality. I, on the other hand, would not practice my arm flapping skills.

And I do believe it is possible and even plausible for AP2 to be ~2x safer than the average human driver. But I suspect achieving that level of safety will require a sufficiently conservative driving style such that it will be unacceptable to all but the most patient human passengers - and all but the most patient of fellow motorists.

 

[disagree]

I @disagree ;-)

Tesla has two very good reasons not to focus on upgrading your car,
- It diverts attention and resources from creating the next new toy. Building in upgradability is a cost.
- It serves as a disincentive to prospective buyers to buy a new car.

History has shown that the only way to upgrade Tesla, is to buy a new one. I don't expect this to change.

I think we actually agree on a significant part of this! Tesla will not want to upgrade a single car.

I just think that the second generation (or 3rd or whatever) of Tesla owners are less likely to forgive Tesla for not providing what the customers feel that they were told they'd have. And Tesla will have to make a decision.

 

[ABC2D]

Maybe there is more to Tesla's progress on FSD that we know. According to this article, while there is no exact timeline (so typical Tesla-like), the deliveries being planned for July.

Tesla signs deal to supply 200 Model S/X vehicles with ‘fully self-driving-capability’ to Dubai’s taxi company

 

[Reeler]

FSD is "dependent on extensive software testing" which says to me that if the software don't work they are off the hook if they expended their best efforts getting it there after several years of trying. Sure, you will get FSD almost all the time meaning 90% of the time which is about the amount of time cars are on the freeway.

Welcome to the perma-Beta world for your production car that is enjoyed in the Tesla reality distortion field. Silicon snake oil.

 

[mhw23]

These cars are like big iPhones. Skip every other generation. I'm waiting for AP3.

 

[ChadFeldheimer]

The delay in getting AP 2.0/EAP to provide at least AP 1.0 functionality is understandable, because they had to adapt the software to use data from different sensors. Remember, it took a year after AP 1.0 hardware was introduced before the software was activated - it's likely AP 2.0 will be activated to AP 1.0 functionality much faster than that - likely in the next few months, though longer than Tesla had predicted.

But it seems like AP2 sensors are a superset of AP1 sensors - so in the worst case, for AP2 Tesla could simply ignore the data from sensors that were not available in AP1 to deliver AP1 features to AP2 customers. I do recall the ultrasonic sensors have additional range in AP2 vs AP1, but this does not seem like a showstopper - applying a filter to translate the AP2 ultrasonic sensor output to AP1 sensor equivalent can't be all that difficult, if even necessary.

Assuming AP2 sensors are a superset of AP1 sensors, AP2 could even have been trained to AP1 performance using the AP1 training dataset - assuming the training data was / is available to Tesla. Mobileye no doubt maintains some private training data, but it's unclear how much of that was actually leveraged by Tesla.

 

[182RG]

who will beat Tesla to market 24 months from now with full autonomy and how?

Driverless Cars: 6 Firms On The Cutting Edge | Bankrate.com

L5? No one. However, Tesla won't be full autonomy in 24 months either. It has to perform flawlessly, and process all edge cases to an astronomically high degree of accuracy.

Audi says 20-30 years, due to the overwhelming task and technology needs.

Stertz says a fully automated vehicle with no driver is still 20 or 30 years away. "To have the car understand every single possibility is a massive challenge," he says.

The rest all say years and decades.

So does EM know and believe something that the rest of the industry doesn't know, or.....?

 

[No2DinosaurFuel]

Everyone who thinks AP2 won't have the stuff to do FSD, then just wait for AP3. That is what I am doing with my AP1. Rock it until AP3 comes along and then maybe I will switch assuming the competitors are still way behind tesla in everything like they are right now.

 

[calisnow]

I think we actually agree on a significant part of this! Tesla will not want to upgrade a single car.

I just think that the second generation (or 3rd or whatever) of Tesla owners are less likely to forgive Tesla for not providing what the customers feel that they were told they'd have. And Tesla will have to make a decision.

Stop playing language games with us. Not in the mood.

 

[disagree]

Stop playing language games with us. Not in the mood.

Is that a riddle?