Here's How Tesla will launch the 'Tesla Network' as L2 ADAS!

[diplomat33]

Which is why they will use "used leased model 3s" to do it. while they are waiting to be sold/re-leased.

Maybe one thing that we can agree on, is that Tesla using post-lease Model 3's in some sort of ride-sharing network actually does make a lot of business sense because it would allow Tesla to make money from cars that would otherwise just sit there. Monetizing cars that would otherwise just sit in a lot doing nothing for you, is smart business. I can think of a lot of car dealerships that would love to do that. So putting aside the question of autonomy, just from a purely business perspective, a Tesla ride-sharing network would be a really big deal.

 

[diplomat33]

It is worth noting that Tesla is only now starting Model 3 leases so it will be a year or two before Tesla even starts to get Model 3's back from leases that they could use in a Tesla Network. So Tesla might be planning ahead that they expect to achieve something close to L4 autonomy by the time the first Model 3 leases end.

 

[OPRCE]

If Tesla demos something that is still far from L4, then scenario 3 will be the most likely

L4 is too far fetched but let's suppose they demo something that with 6 months further refinement could pass muster as a safe and reliable L3 for general purpose use up to the posted speed limits on all public roads driven by Google Maps and giving a minimum of 5s warning for driver to resume control:

1. That in itself would be a bombshell development, the first such [mass-produced] vehicle anywhere.
2. There's no uniform legal framework to sell/operate it as >=L3 [proposed federal law is in limbo] so may have to start in California only.
3. California still requires specially-trained safety drivers in each AV test vehicle, so casual members of general public are ruled out.
4. That would mean employees, unless they stick to L2 until regulations finalise and tech ripens, in which case owners are again in play.
5. #4 also applies if the demo is quite underwhelming.

So Tesla might be planning ahead that they expect to achieve something close to L4 autonomy by the time the first Model 3 leases end.

While Musk doubtless can convince himself of anything, "close to L4" won't work though, it has to be actual L4 with full legal approval for the Robo-Taxi vision to launch. Missing the target by the ritual 6..12 mths, potentially leaving thousands of off-lease M3s stranded in parking lots for the Shorty AirForce to gloat over, would be a large financial and PR disaster they are not forced to make by betting heavily on Elon's established shaky ability for FSD forecasts. This is the time to let caution be the better part of valour methinks!

I want to emphasize the point that Model 3's have that internal camera that could monitor driver attention and therefore remove all hands on wheel nags. So the Model 3's in the Tesla Network may have zero hands on wheel nags. In that case, wouldn't that actually be L3, not L2, since there is no hands on wheel?

No, because if it is designed to require any driver attention for all but emergency takeover situations then it is still L2.

 

[Nocturnal]

This is not a bad theory, they certainly might try something like that — anything to save face I guess.

Save face? Is anyone else doing anything remotely similar?

L4 is too far fetched but let's suppose they demo something that with 6 months further refinement could pass muster as a safe and reliable L3 for general purpose use up to the posted speed limits on all public roads driven by Google Maps and giving a minimum of 5s warning for driver to resume control:

1. That in itself would be a bombshell development, the first such [mass-produced] vehicle anywhere.
2. There's no uniform legal framework to sell/operate it as >=L3 [proposed federal law is in limbo] so may have to start in California only.
3. California still requires specially-trained safety drivers in each AV test vehicle, so casual members of general public are ruled out.
4. That would mean employees, unless they stick to L2 until regulations finalise and tech ripens, in which case owners are again in play.
5. #4 also applies if the demo is quite underwhelming.

No nags but an awake driver would be awesome. You could read, check emails, talk to your passengers etc. It would only force you to engage if it couldn't handle something. Road closures, weather, hardware failure etc. Definitely a game changer even at just L3.

 

[electronblue]

Save face? Is anyone else doing anything remotely similar?

My reference to saving face was related to Tesla’s announcements in 2016 regarding FSD and Level 5 capable hardware. Re-formulating it as a Level 2 Tesla Network certainly would help remedy that image, even though it is not quite what people would have expected, if delivering a Level 5 Tesla Network proves to be too hard at this time.

But of course as Tesla has not announced any Level 2 Tesla Network all this is just interesting speculation.

 

[scaesare]

... it relies on untrained drivers who ain't familiar with Teslas...

AKA "drivers". What happens with ZipCar and other car-sharing services people use when they get in to a car they haven't driven before... are you suggesting they are befuddled and would be incapable of operating that car safely?

...it will drive with you as the driver and will alert you to be fully prepared to take over at any time...

And this is a problem because that steering wheel and brake pedal thingy will be such a foreign concept to them?

But there will be casualties.

As there would be if they drove their own car. Or walked. Or grabbed a Bird scooter they were "unfamiliar with".

The threshold is not "0 casualties". The threshold is simply "better than non assisted".

 

[electronblue]

@scaesare

I will be the first to admit this idea is ”out there” but it is at least one idea to answering how Tesla Network might lauch in the short-term when expecting Level 4/5 driving is probably too much. A Level 2 Tesla Network could be done by end of year belieavably...

What this type of Tesla Network could offer is geofencing and billing that would protect your rented car by forcing the general route taken and ensuring billing to match while the rental driver would still have responsibility.

Another alternative was suggested that Tesla could do this with safety drivers in-car and develop their autonomous tech step by step while already operating the Network.

The other alternatives that makes any sense is that Tesla is far from actually launching the Network. Or that they are very close to launching Level 4-5 driving which seems a bit a of reach too.

 

[scaesare]

@scaesare

I will be the first to admit this idea is ”out there” but it is at least one idea to answering how Tesla Network might lauch in the short-term when expecting Level 4/5 driving is probably too much. A Level 2 Tesla Network could be done by end of year belieavably...

What this type of Tesla Network could offer is geofencing and billing that would protect your rented car by forcing the general route taken and ensuring billing to match while the rental driver would still have responsibility.

Another alternative was suggested that Tesla could do this with safety drivers in-car and develop their autonomous tech step by step while already operating the Network.

The other alternatives that makes any sense is that Tesla is far from actually launching the Network. Or that they are very close to launching Level 4-5 driving which seems a bit a of reach too.

I have no idea if this is the plan or not either.

I'm attempting to provide a little context regarding the "worry language" @Bladerskb opts to use in his description that tends to paint the situation as reckless.

 

[shrfu31]

Tesla is holding an autonomy event April 22. It was rescheduled from the 19th. They also bumped their earnings report date forward.

No other manufacturer is hosting an autonomy event that I know of.

I can’t even begin to speculate that this is something negative for the company. Why would they engage in an unnecessary self-own? You have to be incredibly imaginative to somehow conclude such a pessimistic view of the event.

Maybe I will be wrong but I am not going to pass any judgments much less carry my opinion as fact until I see the event myself. It’s a week away, how cool is that?

 

[malcolm]

Wow Blader, you allowed Musk L2?

Thought you were going to suggest Level zero - build a few more tunnels between fixed locations over the next few years while the M3 leases run down. Then just run them through the tunnels on Boring Company skateboards.

 

[diplomat33]

Here is what Musk said about the price structure of the Tesla Network:
"This was something that Elon Musk himself highlighted in the past. “We would charge something comparable to how you’d say the App Store works, or I don’t know, we’d charge 30% or something in order for somebody to add the car to the fleet. I think that’s a pretty sensible way to go,” Musk said." Tesla’s inclusion of Autopilot on every car unravels a long-term Full Self-Driving strategy

Does this mean that Tesla would take a 30% commission from every ride and the owner of the car would get 70%?

Tesla has now made basic AP standard on all Tesla cars going forward. They are putting the AP3 computer, dubbed "full self-driving computer" in all cars. They are hosting an autonomy day on April 22 where they will apparently do test drives of FSD. And they have announced that leased Model 3 cars will be used for a full autonomy ride-sharing network. Tesla seems to be going all in on full autonomy. Certainly, Tesla seems to believe that they will have full autonomy at some point. Like I said before, if Tesla is bluffing, it's one heck of a bluff because they are certainly acting very confident about full autonomy.

 

[Axtrader]

This all sounds like a preplanned response by TESLA to the Uber IPO Announcement and a means to catch that wave.

 

[TMeister]

It is worth noting that Tesla is only now starting Model 3 leases so it will be a year or two before Tesla even starts to get Model 3's back from leases that they could use in a Tesla Network. So Tesla might be planning ahead that they expect to achieve something close to L4 autonomy by the time the first Model 3 leases end.

Agreed.

I'm expecting that Tesla wants to own these lease returns because they will fit them with a new sensor suite and then test them in the Tesla Network. That keeps control with Tesla (only lease returns, not customer cars) and probably allows a reasonable time frame for something better than L2. I'm not saying L4 since I think we are decades away not years away from that.

There's gotta be a thread somewhere here on why the academics think FSD is not solvable in the foreseeable future while manufacturers keep telling us that FSD is right around the corner. The discussion seems to be mixed in with other threads.

 

[S4WRXTTCS]

Agreed.

I'm expecting that Tesla wants to own these lease returns because they will fit them with a new sensor suite and then test them in the Tesla Network. That keeps control with Tesla (only lease returns, not customer cars) and probably allows a reasonable time frame for something better than L2. I'm not saying L4 since I think we are decades away not years away from that.

There's gotta be a thread somewhere here on why the academics think FSD is not solvable in the foreseeable future while manufacturers keep telling us that FSD is right around the corner. The discussion seems to be mixed in with other threads.

We're likely decades away from a full roll-out of L4 over all kinds of areas.

But, I think we're a lot closer to L4 in geo-fenced areas than people might believe.

What's really holding things up is largely fear. That fear is the question of what's level of safety is required by autonomous vehicles.

When I look at Tesla sensor suite I don't see the kind of redundancy that I'd expect to see in a L3 vehicle let alone an L4 vehicle. In fact I'm amazed that Tesla enabled ULC. That's an awful lot of trust in humans to double check before allowing the vehicle to go over.

So while the Tesla does lack the redundancy really needed what they don't lack is the courage/stupidity/etc to attempt FSD. Even to the point of it only being a semi-quasi L2/FSD mix.

It's so completely insane that it might just work. Not in an immediate "out-of-the-box" work, but through iterations or "mistakes".

Will there be accidents? Yes
Will there be say many "humans can't expect to maintain vigilance while something else is driving" arguments that it will make us all go insane? Yes
Will there be fatalities connected? Yes
Will some states/cities try to ban this weird Tesla experiment? Probably rumbles of it

Sometimes in life you carefully plan things out like an academic. Where it never goes anywhere except for a paper or two.

Other times in life you just go for it.

Tesla is just going to go all in with all guns blazing.

I'm on the ride because there is no other ride.

 

[electronblue]

Agreed.

I'm expecting that Tesla wants to own these lease returns because they will fit them with a new sensor suite and then test them in the Tesla Network. That keeps control with Tesla (only lease returns, not customer cars) and probably allows a reasonable time frame for something better than L2. I'm not saying L4 since I think we are decades away not years away from that.

There's gotta be a thread somewhere here on why the academics think FSD is not solvable in the foreseeable future while manufacturers keep telling us that FSD is right around the corner. The discussion seems to be mixed in with other threads.

On the other hand I would not be surprised if Tesla does not keep any of the leased vehicles in the end. They have years to come up with a plan and in the end it may be a different plan. The lease move gives them options but does not preclude selling the cars after the lease. It is possible the lease thing is just Tesla Network PR and a way of buying time — it could possibly also be means of not commiting or communicating a residual, if they find that lucrative from some financing perspective for Tesla.

 

[Reciprocity]

True but Its all about the Hype/Marketing/Branding. So it will just be a L2 ADAS but be marketed as "full self-drivinf" and the hyping, marketing and branding of Elon/Tesla would make it seem "revolutionary".



Its interesting that you got disagreed on that post. shows you that back then saying FSD was L2 would face stiff resistance because people expect FSD to launch in 2018 with the Tesla Network.



Its always funny to watch you flip flop when it suites you.

You should know, I don't read your responses, not starting today.

 

[CarlK]

We're likely decades away from a full roll-out of L4 over all kinds of areas.

But, I think we're a lot closer to L4 in geo-fenced areas than people might believe.

What's really holding things up is largely fear. That fear is the question of what's level of safety is required by autonomous vehicles.

When I look at Tesla sensor suite I don't see the kind of redundancy that I'd expect to see in a L3 vehicle let alone an L4 vehicle. In fact I'm amazed that Tesla enabled ULC. That's an awful lot of trust in humans to double check before allowing the vehicle to go over.

So while the Tesla does lack the redundancy really needed what they don't lack is the courage/stupidity/etc to attempt FSD. Even to the point of it only being a semi-quasi L2/FSD mix.

It's so completely insane that it might just work. Not in an immediate "out-of-the-box" work, but through iterations or "mistakes".

Will there be accidents? Yes
Will there be say many "humans can't expect to maintain vigilance while something else is driving" arguments that it will make us all go insane? Yes
Will there be fatalities connected? Yes
Will some states/cities try to ban this weird Tesla experiment? Probably rumbles of it

Sometimes in life you carefully plan things out like an academic. Where it never goes anywhere except for a paper or two.

Other times in life you just go for it.

Tesla is just going to go all in with all guns blazing.

I'm on the ride because there is no other ride.

Tesla is going full throttle (pedal?) for it. If people cite accident rate of Tesla FSD cars Elon will counter-cite accident rate of human driven cars. Which one you should ban if the purpose is to make driving safer? It's pretty clear that is his strategy. The target is 2x~3x statistically significant safer FSD cars and the war will be on.

 

[diplomat33]

Tesla is going full throttle for it. If people cite accident rate of Tesla FSD cars Elon will counter-cite accident rate of human driven cars. Which one you should ban if the purpose is to make driving safer. It's pretty clear that is his strategy. The target is 2x~3x statistically significant safer FSD cars and the war will be on.

Yes and I would argue that even 2x or 3x safer would save a lot of lives each year. I know this thread gets caught up in arguments about is L2 or L4 autonomous but really any system that can be 2x or 3x safer than human drivers would be a great achievement in itself.

 

[S4WRXTTCS]

Tesla is going full throttle (pedal?) for it. If people cite accident rate of Tesla FSD cars Elon will counter-cite accident rate of human driven cars. Which one you should ban if the purpose is to make driving safer? It's pretty clear that is his strategy. The target is 2x~3x statistically significant safer FSD cars and the war will be on.

Ultimately Elon can't dictate what society will accept when it comes to L3/L4/L5 autonomous driving.

There is also wide ranging differences in human accident rates depending on the region, vehicle type, etc. It's also well known that bad human drivers account for a large amount of the accidents. The people who have drug/alcohol problems, self control problems, etc.

The truth is that most human accidents really aren't accidents at all, but are the direct result of a willful disregard to rules of the road.

How can those crash incidents possibly be used to prove the safety of autonomous cars?

Autonomous cars have to be 2x better than the best human drivers.

Odds are the better drivers will be those that adopt self-driving cars first. We're likely older, more educated, and have more stable lives.

So why in world would we engage L3/L4 (as in unsupervised) autonomous driving on a car that wasn't significantly better at driving than ourselves?

I don't see that happening, and instead it's going to be a long process of significant improvements (what I expect to see with HW3) followed by a long period of incremental improvements. Where certain road types (freeways) will be allowed for L3 driving a lot sooner than other areas.

The most important role Tesla has in the evolution of autonomous driving is actually the darkest role.

Right now humans don't accept autonomous driving deaths, and we'd never let them on the road in any real capacity regardless of how many lives they saved. Humans aren't very logical when it comes to trolley car problems.

The genius of Tesla's approach is it allows improvements to autonomous driving on a large scale while also still being able to blame the human for an accident. Even when the accident really was the fault of the autonomous car. The mass market approach to it will push people to not only accept autonomous driving, but to want it.

It will save lives in the long run as it will get us to autonomous driving sooner. It puts pressure on other companies to match/exceed what Tesla will offer.

Sure I'd like to see Tesla at a better starting point where they had things like rear corner radars, and down facing 360 degree cameras. But, hopefully what they have is enough to get this started.

 

[DCEV]

@S4WRXTTCS

I don’t disagree with anything you say there.

One more thought: What if you don’t think about Tesla Network so much as an autonomous driving network but as a built-in geofence and billing network. In this speculation the car would allow and indeed require limited driving from the driver but at the same time control where the car goes and how much is billed for it. Weather would not matter that much here, driver could just take over as needed but the car would refuse to stray too far.

I could actually see Tesla lauching this feature to owners in a not too distant future. You can place your car in the network, Tesla takes a share, people drive your semi-autonomous car and the software keeps you car relatively safe and the money flowing. It is not bad speculation.

Didn't you just describe Turo?