What if FSD doesn't materialize?

[Goose66]

It seems more and more likely that there may come a point where they will have to concede that 8-camera vision and ultrasonics just can't do L5 (or maybe even L3). Then what? Will they offer a refund? Will they offer LIDAR, RADAR, and/or HW4 upgrades? Will they just declare victory with whatever it can do? Will they just keep "working on it" in perpetuity in order to never admit defeat?

I have to believe end of 2021 is pretty well set in stone at this point (and yes I realize many end-of-year delivery promises for FSD have come and gone since 2016). I guess will just have to see what happens.

 

[jjrandorin]

Note: I am not a moderator for this section, and making this post as just a regular user.

Why is this a separate thread, instead of in one of the many "FSD" discussion threads, like:

After what time has passed would you consider an FSD class action lawsuit?

And what type of reimbursement would you expect? By FSD, I mean full hands off autonomy & sleeping in your car. Not beta.

teslamotorsclub.com

or

Autonomous Car Progress

This thread is just a stub to continue discussion on autonomous car progress by other manufacturers and avoid the continual hijacking of other useful threads to discuss this subject.

teslamotorsclub.com

or even

What is SAE Level 5 and can Tesla actually achieve it?

Many people, including myself, have strong opinions on level 5, but few have attempted to read and understand the SAE definitions. Here, we discuss whether or not Tesla can achieve level 5 in the near future, or is it impossible (given the sensor suite and/or feature limitations)? To create a...

teslamotorsclub.com

which are still on the first page of this forum?

 

[Goose66]

Why is this a separate thread, instead of in one of the many "FSD" discussion threads, like...

Because none of those threads address the question posed? If the post is a violation of some rule, then please move or delete it as necessary.

 

[Fred42]

It seems more and more likely that there may come a point where they will have to concede that 8-camera vision and ultrasonics just can't do L5 (or maybe even L3). Then what? Will they offer a refund? Will they offer LIDAR, RADAR, and/or HW4 upgrades? Will they just declare victory with whatever it can do? Will they just keep "working on it" in perpetuity in order to never admit defeat?

I have to believe end of 2021 is pretty well set in stone at this point (and yes I realize many end-of-year delivery promises for FSD have come and gone since 2016). I guess will just have to see what happens.

Your Tesla contract doesn't include a promise to accomplish level 5, so I don't envision refunds. Tesla will keep making improvements, and it will get better and better. I'm hoping that because the EU will require a hands-free driver monitoring system that Tesla will implement this in the US as well, and if so I'll be satisfied even if they don't make it to level 5 before 2025 or so, which is my expectation.

 

[DanCar]

> Will they offer LIDAR
Won't help. I believe Elon with his aggressive approach (ability to take risk), will likely be first with broad deployment before LIDAR crowd.

> , RADAR
Tesla has that.

> HW4 upgrades?
I think that is likely. About two years ago during Tesla autonomy day, Elon said HW4 will come out in about two years.

Problem is not the hardware. Problem is the software. Eventually we will get there, but it will take much longer than people hoped. Suspect we will have some false starts also. Where something is rolled out, then has to be rolled back, due to unforeseen circumstances.

 

[S4WRXTTCS]

My guess is it will be just like everything else Elon does where its continuous cycle of try, fail, change things up, fail, etc.

The only thing they have to achieve is getting FSD good enough as an L2 system to sell it as a subscription, and stop selling it as a product.

Once they have it as a subscription model with reasonably good performance in key areas they can increment the hardware to make it better for a new vehicle. Will old customers be mad? yeah, but this hasn't stopped Tesla before. They'll probably have some kind of trade in deal for people with FSD where they get X years of free FSD on the new vehicle.

Now this approach is only applicable for regions that allow L2 FSD as it currently exist in the Beta.

 

[S4WRXTTCS]

> , RADAR
Tesla has that.

My expectation is they'll upgrade the front Radar, and they'll add side/rear radar.

They currently have no redundancy when doing a lane change unless you could the ultrasonic sensors.

 

[Tam]

It seems more and more likely that there may come a point where they will have to concede that 8-camera vision and ultrasonics just can't do L5 (or maybe even L3). Then what?

You need to differentiate between salesman talk vs what you bought when you click the button to buy Autopilot/EAD/FSD.

Will they offer a refund?

No. There's a language explained what you bought so if you don't agree, you don't click to buy

Will they offer LIDAR, RADAR, and/or HW4 upgrades?

They could but with an additional charge or you might have to pay for a whole new car with HW upgrades.

Will they just declare victory with whatever it can do?

Already did. It's just "smooth like silk", "amazing"...

Will they just keep "working on it" in perpetuity in order to never admit defeat?

Why not? That's what you agreed when you read what you clicked to buy

I have to believe end of 2021 is pretty well set in stone at this point

Good for you.

I guess will just have to see what happens.

That sounds like an eternity.

 

[Kingoglow]

I think (at least) 2 more cameras are needed. There are several videos out illustrating why the B-Pillar cameras are not good enough to see to the left and right of the vehicle at some intersections. Two more, side-facing cameras near the headlights would help in scenarios where vision is needed near the front bumper of the car when the car is in danger of creeping into traffic to establish vision.

While this video doesn't talk about the issue, there are several instances where you can imagine that having cameras more far forward would help.
Chuck Cook - FSDBeta 8.2 - 2021.4.11.1 - Left Unprotected Turns with Drone 3rd Person View - 4K

HW4 (and beyond) is defenitly coming.

 

[Goose66]

My expectation is they'll upgrade the front Radar, and they'll add side/rear radar.

They currently have no redundancy when doing a lane change unless you could the ultrasonic sensors.

Yes, that's what I was referring to adding additional radar - e.g., 3D front radar and side radar in front bumper for unprotected turns.

 

[Goose66]

For those replying with the "read the fine print" for refund, remember that conceding defeat without refunds or some other remuneration could be a big PR blow that could seriously affect the stock price (at least in the short term).

As I said, it seems the latest round of promises has the end of 2021 date well entrenched, and if they don't have it by then they will be faced with a decision. However, there is also a good possibility that regulators will come to their rescue by saying they have to halt the current Beta rollout and marketing strategy before that date, and that would give them cover for why they were unable to fulfill the latest promise.

 

[Tam]

...that would give them cover for why they were unable to fulfill the latest promise.

Now you got it! There's a footnote saying how owners can get what they want and why they won't. One of those is regulators. If some governments don't allow it then the consumers' money goes all down the drain.

 

[run-the-joules]

As I said, it seems the latest round of promises has the end of 2021 date well entrenched, and if they don't have it by then they will be faced with a decision.

The order page itself said autosteer on city streets by end of 2019, and then by end of 2020, and all they did was update the order page after blowing those timelines. There's nothing different about this year.

 

[Goose66]

The order page itself said autosteer on city streets by end of 2019, and then by end of 2020, and all they did was update the order page after blowing those timelines. There's nothing different about this year.

I guess I am going more by Elon's tweets, which have gone from "on track for" "feature-complete" by end of 2019 to "confident" of general availability of L5 FSD by the end of 2021. But it's all in the way you read it, I guess.

 

[run-the-joules]

I guess I am going more by Elon's tweets, which have gone from "on track for" "feature-complete" by end of 2019 to "confident" of general availability of L5 FSD by the end of 2021. But it's all in the way you read it, I guess.

Earnings call - Feb 19 2019: "We will be feature complete full self driving this year. The car will be able to find you in a parking lot, pick you up, take you all the way to your destination without an intervention this year. *****I'm certain of that. That is not a question mark*****. It will be essentially safe to fall asleep and wake up at their destination towards the end of next year"

 

[Dan D.]

However, there is also a good possibility that regulators will come to their [Tesla's] rescue by saying they have to halt the current Beta rollout and marketing strategy before that date, and that would give them cover for why they were unable to fulfill the latest promise.

The only reason regulators might step in is if they feel the product is not safe. That is all on Tesla. Make a safe product and obey the rules in releasing it; that is all the regulators want. Don't blame regulations if you don't follow them. I wouldn't accept that as an excuse.

 

[Matias]

Now you got it! There's a footnote saying how owners can get what they want and why they won't. One of those is regulators. If some governments don't allow it then the consumers' money goes all down the drain.

Tesla has not tried to get regulatory approval, so that is not applicable.

 

[FloridaJohn]

Tesla has not tried to get regulatory approval, so that is not applicable.

Where do they apply for regulatory approval?

 

[JHCCAZ]

I think (at least) 2 more cameras are needed. There are several videos out illustrating why the B-Pillar cameras are not good enough to see to the left and right of the vehicle at some intersections. Two more, side-facing cameras near the headlights would help in scenarios where vision is needed near the front bumper of the car when the car is in danger of creeping into traffic to establish vision.

While this video doesn't talk about the issue, there are several instances where you can imagine that having cameras more far forward would help.
Chuck Cook - FSDBeta 8.2 - 2021.4.11.1 - Left Unprotected Turns with Drone 3rd Person View - 4K

HW4 (and beyond) is defenitly coming.

Thanks for raising this point. I've spent a good while watching FSD beta videos, and to me the Number One mystery is why they did not place left- and right-looking cameras much farther forward - around the headlights or at least in the A-pillars - to give cross-traffic vision angles that are better than the human driver has, rather than worse. And this point only gets amplified if another car pulls alongside you while you're waiting to turn.

So I do think such forward-mounted cameras could greatly improve the confidence in initiating turns into cross-traffic, with less of the hesitant creeping-out behavior. However, I'd say that Chuck Cook's Important demonstration, of the current Beta-s poor timing and decision-making for unprotected lefts across oncoming traffic, would not be helped much by more forward-mounted cameras. I think that's more attributable to presently-inadequate consideration of the time dimension.

Human drivers see the left-turn opportunity coming up, make a decision and initiate the turn just as the last oncoming vehicle is passing by. Chuck's video shows the Tesla waiting too long, debating with itself in the moment (thus missing the safest moment) and then sometimes beginning a late and dangerous turn in front of approaching cars. All this is bad, but I don't see it as a problem of poor camera angles.

There's been a fair amount of discussion that "FSD 9" will have much better temporal extrapolation and movement prediction for tracked objects, including for example the expected re-appearance of vehicles or people that disappeared behind some obstruction. My hope is that this general capability will contribute to far more human-like predictive skills for picking the right moment to initiate a smooth and confident turn.

 

[stopcrazypp]

Thanks for raising this point. I've spent a good while watching FSD beta videos, and to me the Number One mystery is why they did not place left- and right-looking cameras much farther forward - around the headlights or at least in the A-pillars - to give cross-traffic vision angles that are better than the human driver has, rather than worse. And this point only gets amplified if another car pulls alongside you while you're waiting to turn.

Even without reference to those videos, having bought a Model 3, that was one of the things I noticed too. Not just for FSD, but cameras mounted more forward and lower would help enable the 360 parking views that many other cars have today.

So I do think such forward-mounted cameras could greatly improve the confidence in initiating turns into cross-traffic, with less of the hesitant creeping-out behavior. However, I'd say that Chuck Cook's Important demonstration, of the current Beta-s poor timing and decision-making for unprotected lefts across oncoming traffic, would not be helped much by more forward-mounted cameras. I think that's more attributable to presently-inadequate consideration of the time dimension.

Human drivers see the left-turn opportunity coming up, make a decision and initiate the turn just as the last oncoming vehicle is passing by. Chuck's video shows the Tesla waiting too long, debating with itself in the moment (thus missing the safest moment) and then sometimes beginning a late and dangerous turn in front of approaching cars. All this is bad, but I don't see it as a problem of poor camera angles.

There's been a fair amount of discussion that "FSD 9" will have much better temporal extrapolation and movement prediction for tracked objects, including for example the expected re-appearance of vehicles or people that disappeared behind some obstruction. My hope is that this general capability will contribute to far more human-like predictive skills for picking the right moment to initiate a smooth and confident turn.

That video, even in the tries that failed, the cars in the opposing lane were marked yellow (the color for oncoming vehicles) and were detected successfully, so perception/sensors were not the issue. It's the AI that makes the decision to go or not that is the problem.