HW4 sensor suite and hi-res camera updates?

[Mardak]

Basically is it worthwhile for Tesla to assume responsibility for L3

Oh indeed, it seems like conditional driving automation at least so far is relatively limited in profit for automakers while carrying a lot of potential costs unless extremely limiting the scope which then limits the potential value that consumers would pay for the feature.

People have been asking for "eyes off," but Tesla and many other automakers don't seem that interested.

 

[lowtek]

I did not. I specifically paid for a car with FSD, which is why I made sure to put it on my original order, not tacked on after delivery. I also paid for 20" wheels and 7 seats in my original order, and got them. If they were backordered, that's fine. But if they fail to eventually deliver in reasonable time, or rather, deliver to others that bought after me, and not to me, then I'd use legal action to enforce the order, take back the car, or pay damages.

This isn't difficult.

You have to add the "option" specifically, and if you did you got FSD in it's current form with improvements and updates as they become available ... just like buying an app on your phone.

 

[JulienW]

..... Basically is it worthwhile for Tesla to assume responsibility for L3.....

But do they have choice? In Georgia (and I think most states) the party that IS driving the car is first party responsible (not the owner). So if FSD is driving in L3 and hasn't requested a takeover doesn't that automatically make Tesla responsible?

 

[DL_Mike]

You have to add "option" specifically and you did get FSD in it's current form.

View attachment 900024

First of all, it's not optional anymore once I select it in my original order, and pay for it. Technically, buying a Tesla at all is optional, so your logic doesn't really fly.

And as for the screenshot, exactly! If they never get there, fine, but if they get there and have proven feasibility, then they can apply it to my car. If that means they have to give me the hardware to make it work, then they have to do it. They can't then use the defense that they haven't proven feasibility if there are Tesla cars on the road, that aren't mine, that are doing it.

In all reality, for one-off court cases they probably won't even show up. For class actions obviously it's a different story.

 

[JulienW]

Exactly! If they never get there, fine, but if they get there and have proven feasibility, then they can apply it to my car. If that means they have to give me the hardware to make it work, then they have to do it. They can't then use the defense that they haven't proven feasibility if there are Tesla cars on the road, that aren't mine, that are doing it.

In all reality, for one-off court cases they probably won't even show up. For class actions obviously it's a different story.

Again, if (BIG IF) they can get there in HW4 and when you ask for your car they will say they are working to get it for your car (HW3). It is just going to take more time.

Kinda like Parking Sensors features for a new 3/Y. It is coming.

 

[lowtek]

First of all, it's not optional anymore once I select it in my original order, and pay for it. Technically, buying a Tesla at all is optional, so your logic doesn't really fly.

And as for the screenshot, exactly! If they never get there, fine, but if they get there and have proven feasibility, then they can apply it to my car. If that means they have to give me the hardware to make it work, then they have to do it. They can't then use the defense that they haven't proven feasibility if there are Tesla cars on the road, that aren't mine, that are doing it.

In all reality, for one-off court cases they probably won't even show up. For class actions obviously it's a different story.

Not sure that tracks but please let us know how it goes

 

[EVNow]

But do they have choice? In Georgia (and I think most states) the party that IS driving the car is first party responsible (not the owner). So if FSD is driving in L3 and hasn't requested a takeover doesn't that automatically make Tesla responsible?

Right - thats why current gen, IMO, will not be "L3".

Let me put it another way, unless there is competitive pressure for Tesla to declare L3 and assume responsibility, they will not do it.

They could do something in between in a convoluted way ... while still keeping L2/ADAS tag.
- Use Tesla insurance
- No deductibles if FSD is active
- Driver still responsible (legally)

 

[JulienW]

Right - thats why current gen, IMO, will not be "L3".

Let me put it another way, unless there is competitive pressure for Tesla to declare L3 and assume responsibility, they will not do it.

They could do something in between in a convoluted way ... while still keeping L2/ADAS tag.
- Use Tesla insurance
- No deductibles if FSD is active
- Driver still responsible (legally)

Yea, that could make sense. Tesla could stay in FSD Beta/L2 and even for the HW4. Then they could then make their detected Robotaxi they are likely to announce in March (??) to L4 on HW4 as a commercial only vehicle. Might make a consumer version but keep it at L2 Beta.

 

[Daniel in SD]

Yea, that could make sense. Tesla could stay in FSD Beta/L2 and even for the HW4. Then they could then make their detected Robotaxi they are likely to announce in March (??) to L4 on HW4 as a commercial only vehicle. Might make a consumer version but keep it at L2 Beta.

haha, there is zero chance that Elon wouldn’t say something exposing Tesla to the liability they were trying to avoid.

 

[JulienW]

Dichotomy

haha, there is zero chance that Elon wouldn’t say something exposing Tesla to the liability they were trying to avoid.

True since Elon is his own dichotomy. Tesla's greatest asset and biggest liability.

 

[EVNow]

Yea, that could make sense. Tesla could stay in FSD Beta/L2 and even for the HW4. Then they could then make their detected Robotaxi they are likely to announce in March (??) to L4 on HW4 as a commercial only vehicle. Might make a consumer version but keep it at L2 Beta.

The reason I say no L3 with current gen is ... I don't think SFDb will be close to L3 (let alone L4) anytime soon. By the time they get to L3/L4 (atleast 100x better than now in terms of disengagement rate), say in the proverbial 5 years, the current gen vehicles would be in a minority. We might be looking at Gen 3 platform vehicles and HW5/6 by then.

 

[spacecoin]

Right - thats why current gen, IMO, will not be "L3".

I agree that HW3 won't ever become anything else than L2. But the reason is that the system is at a less than 10 miles MTBF (after 25 months of public "testing"), and you need at least 2000x that to claim autonomy - probably more. If they reach the required level of reliability, a liability insurance will be cheap and not an issue.

 

[EVNow]

But the reason is that the system is at a less than 10 miles MTBF (after 25 months of public "testing"), and you need at least 2000x that to claim autonomy - probably more. If they reach the required level of reliability, a liability insurance will be cheap and not an issue.

Yes, I've been posting this now for months.

BTW, I'd not term disengagements as MTBF directly, but they are definitely correlated. I think a disengagement rate of atleast 1 in 1,000 miles is required for L3 and 10,000 for robotaxi. Waymo is around that.

 

[spacecoin]

BTW, I'd not term disengagements as MTBF directly, but they are definitely correlated. I think a disengagement rate of atleast 1 in 1,000 miles is required for L3 and 10,000 for robotaxi. Waymo is around that.

Waymo is at 30k-40k miles per DE and so is Cruise, if one should believe the self-reporting to CA DMV. Waymo was lower (10k) in 2021 because of their transition to the new platform and car. In 2020 they were at 30k.

Also keep in mind that interventions are more frequent than the DE:s on the beta tracker. Perhaps once per 1-2 miles.

 

[EVNow]

Waymo is at 30k-40k miles per DE and so is Cruise, if one should believe the self-reporting to CA DMV. Waymo was lower (10k) in 2021 because of their transition to the new platform and car.

Also keep in mind that interventions are more frequent than the DE:s on the beta tracker. Perhaps once per 1-2 miles.

Yes, I'm using their 2021 numbers. Not that it matters that much. But we know 1 in 10k is good enough to run robotaxis.

During the year, Waymo’s vehicles disengaged a total of 292 times, for a rate of 0.126 disengagements for every 1,000 miles, or 1 disengagement for every 7,200 miles. That’s a worse rate than it had in 2020, in which the company reported only 21 disengagements over the course of the year, or a rate of 0.033 disengagements per 1,000 miles. The company says it experienced a higher number of disengagements because it introduced a new vehicle to its fleet: the Jaguar I-Pace electric SUV.

Interventions are a bit more vague in FSDb. For eg., I intervene to change max speed setting all the time. If they introduce separate speed setting limits for marked and unmarked roads, the # of interventions would suddenly plunge by 90%.

 

[spacecoin]

Yes, I'm using their 2021 numbers. Not that it matters that much. But we know 1 in 10k is good enough to run robotaxis.

1 in 10k may be enough to run robotaxis if you have teleops and a small fleet. But again, hw3 is not capable of getting to even a 1/10 of that.

If Teslas are autonomous in the LVCC (or a small ODD like it) in three years (on any hardware), I'll be very surprised. I don't think they're even trying for reliability, and ingress/egress is by itself very hard (if one wants to taxi).

If Tesla ever defines an ODD they are dead. If they just keeps going for "real world AI", they can keep this charade going for a few more years.

I wonder when people will stop believing in camera only. 2026? To me it's incredible that people actually believe in L5, regardless of sensing and compute.

 

[EVNow]

I wonder when people will stop believing in camera only.

When they realize humans use Lidars to navigate

 

[pdx_m3s]

This thread has veered way off course.

 

[Dewg]

I did not. I specifically paid for a car with FSD, which is why I made sure to put it on my original order, not tacked on after delivery. I also paid for 20" wheels and 7 seats in my original order, and got them. If they were backordered, that's fine. But if they fail to eventually deliver in reasonable time, or rather, deliver to others that bought after me, and not to me, then I'd use legal action to enforce the order, take back the car, or pay damages.

This isn't difficult.

You're absolutely correct, it's not difficult. Where you're running into trouble is what a court will rule on. Courts work on reason - what is reasonable. Your car is working fine, and has value which you've received. The court will not order Tesla to buy back a car that is working reasonably well. It's why we have Lemon Laws in many states in the US, which can force a company to buy back a car that isn't working reasonably well.

Here's an example:

Car - $100,000
Tires Upgrade (optional) - $2,000
Seats (7-row optional) - $5,000
FSD Package (optional) - $15,000

If the tires weren't delivered, nor the seats, you would know this at delivery and could refuse the delivery, or have a signed document indicating, as you said, there was some parts shortage and you'll get those items at some point in the future. If they never show up, a court would very likely grant you a refund of $7,000 for the tires and seats, plus possible interest on the time.

The same for FSD. If you argue to a court that FSD was not delivered timely, the court would likely grant you a refund of $15,000 for the FSD Package, plus possible interest on the time.

If you could prove to the court that the only reason you purchased a Tesla was for autonomous driving, and that might be difficult to prove, the court would likely grant you a refund on the FSD Package, and possibly the difference between the price of the car ($100,000) and the FMV (fair market value) at the time of the case (let's say KBB shows your car worth $80,000 at the time of court), meaning you'd be granted an additional $20,000. You could then sell your car to be made whole.

Think reasonable.

 

[spacecoin]

When they realize humans use Lidars to navigate

Rather when they realize that computer vision and machine learning is not a human brain and doesn't have common sense and can't adapt to long tail events. Plus, you know human eyes, eyelids, sunglasses, caps and a turning head vs monocular pixel heaps.

Elon need to filter out references to Moravec's paradox from Twitter.