FSD may require a hardware upgrade...

[calisnow]

Elon at press conference: "So all the car are shipping with hardware capable of Level 5."
Elon later that afternoon back at HQ: "Hey, uh, guys. So um, I was on a roll and feeling confident and I may have said 'Level 5' in public. So, erm, yeah - that's what folks expect. Good luck - I know you can do it. Your jobs and my reputation depend on it."
:Elon drops his Diet Coke and runs out the back door before the nearest engineer can strangle him to death:

 

[calisnow]

Not me. I don't have AP2 hardware I do think this is going to blow up in Tesla's face as the average consumer who doesn't read Elon's tweaks or hang out in this forum will be asking why they paid $8K for something that doesn't work after a reasonably amount of time.

Lol oh - I thought you had an AP 2 car and were growing agitated lol. You have a point though, you do have a point.

 

[stopcrazypp]

Not me. I don't have AP2 hardware I do think this is going to blow up in Tesla's face as the average consumer who doesn't read Elon's tweaks or hang out in this forum will be asking why they paid $8K for something that doesn't work after a reasonably amount of time.

On the other hand, your average joe will not be scrutinizing Elon's press conference to find that level 5 reference and about whether whatever is released is technically Level 5. If Tesla released something that the government doesn't deem legal to operate without a driver (thus remaining level 4), that lets them use the regulatory out in the disclaimer.

The single board configuration is definitely capable of level 4, even by Nvidia's estimations. They call it autochauffeur. Nvidia says it won't be enough for level 5, but Elon seems to be banking on it being enough.
Autonomous Car Development Platform from NVIDIA DRIVE PX2

 

[aronth5]

I'm less concerned over the swappable central board then the possibility the sensor suite (ultrasonic sensors, cameras and radar) will not be adequate for level 5.
I can see Tesla supporting the central board upgrade but when you're talking about a different sensor suite that is completely different. Does anyone think the sensor suite won't change in the next 2 years?
What that means only time will tell but hardware always changes.

 

[dhanson865]

Please tell me where it says there that it will be 2018 or 2020 before I get the FSD that was demoed in the video?

It doesn't, now tell me where in the video, or anywhere on www.tesla.com/autopilot does it say it'll be done to the level of that video on any time frame? Big clue is it doesn't. Not days, not months, not weeks, not years. It just says they are working on it and the hardware is in cars already.

and again they toss this in liberally

Please note that Self-Driving functionality is dependent upon extensive software validation and regulatory approval, which may vary widely by jurisdiction. It is not possible to know exactly when each element of the functionality described above will be available, as this is highly dependent on local regulatory approval.

 

[calisnow]

@stopcrazypp - so the question is - why did Tesla not use a more powerful configuration of PX2? It does seem to fit Musk's entire history of "better faster cheaper" development of everything from rockets to cars. And he has achieved industry-best self steering using the "industry's least powerful" sensor suite. So I'll trust his track record. But I wonder if this mentality is based on squeezing dollars til they bleed - or a philosophy that it's easier to write better software if you're fusing fewer sensors.

 

[calisnow]

I'm less concerned over the swappable central board then the possibility the sensor suite (ultrasonic sensors, cameras and radar) will not be adequate for level 5.
I can see Tesla supporting the central board upgrade but when you're talking about a different sensor suite that is completely different. Does anyone think the sensor suite won't change in the next 2 years?
What that means only time will tell but hardware always changes.

You make good points - nobody knows. As for longevity of this sensor suite - I give it 24 months because that's how long the last one lasted, and 2.5 years after the deployment of AP 1 hardware they are still improving the software (perhaps to its final iteration in the soon to be released 8.1).

 

[Electricfan]

@Electricfan - I'll grant you the point about off-ramp to off-ramp. That was not delivered and I don't see how it safely ever could be because the car can't look behind it more than 16 feet. But the performance is just off the charts amazing and stress-reducing now when used within its design limits. It makes multi-hundred mile drives no more tiring than a 5 minute trip to the supermarket. You must have very very sensitive nerves compared to the rest of us. I suppose you think the 40% claimed accident reduction rate from the NHTSA is false as well right?

I enjoyed using it on a 3000 mile road trip last December. But I didn't buy it for road trips. I traded my 2013 S for a 2015 with AP for one reason - autopilot - to use on my commute to work each day. And its useless for that, because the highways are not smooth and straight enough, and the traffic is too thick. During my commute if i use AP I have to grab the wheel at least once, usually more than once each trip. And that makes it worthless to me. That's just my opinion. It was oversold. I hope a rich person out there who bought a model S for AP is as unhappy with it as I am and sues the crap out of Tesla. I'll have to read up on the 40% number you mentioned, I haven't kept up well enough.

 

[CSFTN]

You make good points - nobody knows. As for longevity of this sensor suite - I give it 24 months because that's how long the last one lasted, and 2.5 years after the deployment of AP 1 hardware they are still improving the software (perhaps to its final iteration in the soon to be released 8.1).

While no one knows (if the sensor suite is adequate or will be upgraded), you do have e.g. George Hotz clearly and repeatedly saying - to paraphrase - "humans drive with just their instincts and 1 set of eyes. Nothing more. If we can do that, a machine should be able to do at least as well, with no more sensors."

That certainly makes quite a bit of sense. (Of course it simplifies the action of the mammalian eyeball and neural circuitry, but that's a digression)

 

[MP3Mike]

@stopcrazypp - so the question is - why did Tesla not use a more powerful configuration of PX2?

Because it is cheaper, and Elon thinks they might be able to get away with it. (Look at what they are able to do with one EyeQ3 processor that MobileEye said wasn't possible.) And if they are wrong a couple years down the road they can just upgrade the AP brain box with the new generation of tech for less than it would cost to put it in now. So you can look at it is a bet with a really good pay off if they can pull it off, otherwise they have to dip in to the profit they added to EAP+FSD to do the upgrade. (I highly doubt that anyone that didn't purchase AP or only purchased EAP will need/get an upgrade. And by then it won't cost $5k/$3k they will have raised the price.)

 

[Electroman]

During my commute if i use AP I have to grab the wheel at least once, usually more than once each trip. And that makes it worthless to me.

Really? having forced to hold the wheel once or twice during your commute makes you angry? want to sue tesla, so that you can go back to your Mercedes and hold the wheel forever?

You are the kind of guy who loves to find some fault, anything, imaginary or otherwise. AP1 is just tailor made for highway commutes.

I am sure there is a way to sue people, who sue others just for fun and waste court's time and resources.

 

[dhanson865]

While no one knows (if the sensor suite is adequate or will be upgraded), you do have e.g. George Hotz clearly and repeatedly saying - to paraphrase - "humans drive with just their instincts and 1 set of eyes. Nothing more. If we can do that, a machine should be able to do at least as well, with no more sensors."

That certainly makes quite a bit of sense. (Of course it simplifies the action of the mammalian eyeball and neural circuitry, but that's a digression)

Add the ears and it'd be more accurate. We use both the hearing for clues about things we can't watch non stop and the inner ear ability to judge G forces. Deaf people do drive but most of us use hearing as a cue for if traffic is heavy, if a car is passing us, and so on.

 

[brianman]

He knows he's moving to a whole new market with Model 3 - a less forgiving market. I have 7 personal friends with Model 3 reservations. If he tells them "level 5" - it better f*cking work. You don't get to f*ck that crowd over with bullsh*t and snakeoil and then tell them to go buy a new car.

What does this have to do with the concern of Model S owners that HW2 might be insufficient for FSD?

 

[jbcarioca]

Tesla may have "baked-in" the cost of an upgrade, since it will probably be cheaper to upgrade to newer faster CPU/GPUs once the software is ready than to put in the ones that are fast enough now.

Given the explicit promises made, I'll wager that hardware upgrades, if required, will be done without additional charge if the car has the FSD option already. Almost all of us, I think, believe and have been ed to believe, that FSD= SAE level 4 or 5. I think we'll have little ground for complaint if it becomes Level 4 rather than Level 5, but nothing below Level 4 could be described as FSD.

 

[GSP]

Although Elon said "level 5" in the press conference, the description of the FSD option on Tesla's website and configurator exactly matches the Society of Automotive Engineers description of "level 4." It would be unreasonable for any customer to expect more than level 4. Personally, I would be fine with this.

GSP

 

[brkaus]

I personally would have purchased for L4 only. But it do feel like more was promised, esp with the ability to add ones car to the autonomous for hire fleet.

 

[bob_p]

I requested clarification from Tesla on Musk's comments about the possibility of a future processor upgrade needed for FSD.

It appears the official position is that Musk's comments were hypothetical and doesn't represent Tesla's official position, that they believe the current AP 2.0 hardware is sufficient to support FSD.

AP 2.0 is a combination of the new sensors and the Drive PX2 processor.

Tesla should have been able to do enough testing on the new sensor suite to have confidence the sensors are sufficient to detect all of the objects needed to support EAP and FSD.

The Drive PX2 processors will be used for real-time object recognition and implementation of the FSD AI engine. Tesla has likely done enough preliminary work with the new sensors and Drive PX2 to have confidence they have enough horsepower to do the object recognition.

The biggest unknown is probably with the AI engine. Will there be enough horsepower with the Drive PX2 to do both the object recognition and run the AI software to a level that is at least as good as a human driver, preferably much better? And the answer to that question may not come until Tesla is much further into the FSD development.

While Musk's comments may be viewed by Tesla as hypothetical and not represent Tesla's official position, there is a possibility (however small) that more processing power may be needed to achieve the goals for FSD. But, assuming Tesla's testing of AP 2.0 is currently showing they have considerable idle processing power, it's also likely way too early for Tesla to know how much processing power will actually be needed for FSD.

And, if a processor upgrade will be needed, that should be easy to do. The performance limitations would likely be entirely on the Drive PX2 board - and replacing that board with a faster NVidia board - or a plug-compatible board using a Tesla custom processor would be easy to do - and likely cost either Tesla or customers a relatively small amount, compared to the overall cost of the car or the expensive battery or motor components.

Tesla has assured us they believe the AP 2.0 hardware in the S 100D we have on order (and could be delivered next week!), and, at least for now, Tesla will not make any commitments related to any hypothetical upgrades that may be needed to achieve FSD.

And, we're willing to accept that. We've had a great experience with Tesla's support for our 4 year old S P85, and if Tesla does discover a year or two from now that the Drive PX2 needs to be replaced, we're confident Tesla will do something reasonable to support all of those customers who purchased the FSD option for their cars.

 

[Reeler]

Put me on the list for the free upgrades that get blind spot detection working. My AP1 sensor suite is not capable of doing it and now that Tesla has tried and failed, I expect the hardware upgrades to make it work.

Add to that list, autopark, red light detection, auto wipers, auto high beams, and the perma-beta summon and lane keep. These features work to varying degrees, but not reliably to be useful or with the beta disclaimer to absolve Tesla of all liability should they kill you or ruin your car. Under the @bob_p logic, I should expect hardware upgrades for those promised features too?

The best predictor of future behavior is what has been done in the past for similar circumstances. Please AP2 folks, LOWER YOUR EXPECTATIONS! That is the only true way to enjoy such an amazing ride.

 

[SeminoleFSU]

The problem with that one is there are no damages. The the single and dual motor, and performance dual motor cars were all advertised with an EPA rated range based on a full charge and they achieve that mileage under the EPA tested conditions.

My P85D was advertised at 253 rated miles. I have very little difficulty achieving that as long as it's not too cold, raining, and I don't exceed the speed limit by 5 mph.

what about the newer cars that actually do have the stated capacity? IMO this means that the degradation would be less on them, so the damages to owners of older cars is we have to charge more often/longer doing more damage to the batts to get the same range. Plus, in a salvage situation our batts would be worth less to someone like a Jason Hughes or other folks who just wanted the raw batteries.. i'm not a lawyer, just throwing out things to think about outside of just hitting EPA #s

 

[u00mem9]

The discussion misses the primary failure made by Tesla here. They set the sensor and computational spec before they had successfully developed the software...even at a proof of concept phase. That should be a WTF. Of course you don't install more expensive hardware than needed.

THAT'S WHY YOU CONFIRM THE CONCEPT IN DEVELOPMENT BEFORE YOU START MASS PRODUCING HARDWARE.

All the rest of the excuse making and second guessing the capability of the as-built cars are just a result of this single failure.

And before anyone says AP1 hardware was in vehicles 1 year before they applied the software, I can guarantee you they had that system working 90% well before oct 2014 MP start. This AP2 situation is dramatically more reckless.