Autopilot Mapping as a product

[Hogfighter]

Did anyone catch this in the earnings call?

When asked about the value of the data being collected by autopilot, Elon punted by saying that they didn't want to announce new products on earnings calls.

This essentially means that their data is going to be on an order of magnitude better than what anyone else has. So if other car companies want to provide autopilot, that they'd either have to (1) do what Tesla is doing by having all of their cars be beta testers, which is unpalatable for most purchasers, or (2) purchase the mapping data from Tesla.

I believe that this is a new product that will be announced once the data is more refined, in a few months. Thoughts?

 

[bh1783]

Did anyone catch this in the earnings call?

When asked about the value of the data being collected by autopilot, Elon punted by saying that they didn't want to announce new products on earnings calls.

This essentially means that their data is going to be on an order of magnitude better than what anyone else has. So if other car companies want to provide autopilot, that they'd either have to (1) do what Tesla is doing by having all of their cars be beta testers, which is unpalatable for most purchasers, or (2) purchase the mapping data from Tesla.

I believe that this is a new product that will be announced once the data is more refined, in a few months. Thoughts?

Nokia's mapping unit was bought by all the German carmakers for $3.1billion Euros earlier this year.
http://www.cnbc.com/2015/08/04/the-reason-german-carmakers-bought-nokias-map-unit-google.html

Maybe the mapping can be a product like the superchargers that Telsa could license out to other manufacturers.

 

[jhm]

Did anyone catch this in the earnings call?

When asked about the value of the data being collected by autopilot, Elon punted by saying that they didn't want to announce new products on earnings calls.

This essentially means that their data is going to be on an order of magnitude better than what anyone else has. So if other car companies want to provide autopilot, that they'd either have to (1) do what Tesla is doing by having all of their cars be beta testers, which is unpalatable for most purchasers, or (2) purchase the mapping data from Tesla.

I believe that this is a new product that will be announced once the data is more refined, in a few months. Thoughts?

I think selling enhanced maps has been in the cards for quite a while. They have a fairly unique opportunity to gather the data and a data center for storage and analysis. But I suspect there are lots of gaps to be filled before having a marketable product. Firstly, they area mapped is restricted to where Model S cars roam. There are probable millions of miles of roads where they have no driving data. Secondly, they may need some sort of distribution software to push the data out in usable form. So this depends on who the customer is for this data. If they are selling to some GPS navigation service, then not much is needed beyond a data feed. But if Tesla wants to deliver a navigation service to end user, then they will have to do a lot of work. The navigation system currenly in the Model S is in my opinion not good enough to sell to the broad market. For example, Google Map gives much better optionality and monitoring of road conditions. For Tesla to sell a navigation service beyond the captive market of Tesla vehicles, they will need to seriously step up the game and offer something superior to Google Maps and any other GPS provider out there. I seriously hope they do this, because right now I think that navigation and routing are the weakest links in Tesla's AP suite. You seriously cannot have an autonomous vehicle making ridiculous routing choices.

So I am quite enthusiastic about Tesla moving in this direction, but I think there is a lot of work to be done before they have a marketable product.

 

[Hogfighter]

I think selling enhanced maps has been in the cards for quite a while. They have a fairly unique opportunity to gather the data and a data center for storage and analysis. But I suspect there are lots of gaps to be filled before having a marketable product. Firstly, they area mapped is restricted to where Model S cars roam. There are probable millions of miles of roads where they have no driving data. Secondly, they may need some sort of distribution software to push the data out in usable form. So this depends on who the customer is for this data. If they are selling to some GPS navigation service, then not much is needed beyond a data feed. But if Tesla wants to deliver a navigation service to end user, then they will have to do a lot of work. The navigation system currenly in the Model S is in my opinion not good enough to sell to the broad market. For example, Google Map gives much better optionality and monitoring of road conditions. For Tesla to sell a navigation service beyond the captive market of Tesla vehicles, they will need to seriously step up the game and offer something superior to Google Maps and any other GPS provider out there. I seriously hope they do this, because right now I think that navigation and routing are the weakest links in Tesla's AP suite. You seriously cannot have an autonomous vehicle making ridiculous routing choices.

So I am quite enthusiastic about Tesla moving in this direction, but I think there is a lot of work to be done before they have a marketable product.

Totally agree. I'm curious as to who the customer would be, whether it would be Google, BMW, or the end user. But it's clear that this data holds value.

Model S customers are much more willing to be beta testers than BMW customers.

 

[StapleGun]

Elon punted by saying that they didn't want to announce new products on earnings calls.

This essentially means that their data is going to be on an order of magnitude better than what anyone else has.

That's quite a speculative leap there. I'm glad to see they are thinking about this if only so they don't need to rely on a third party for data, but I think it's quite early to say definitively that they have better data.

 

[replicant]

The market value of a street mapping service could decrease quickly.

Uber, Google, Apple, HERE (now BMW + Audi + Daimler) are all working on building highly-detailed maps. If there's value in Maps as a Service, Google and Uber will fight hard to partners with manufacturers - who will probably prefer to work with them than from a direct competitor like Tesla). Google is already signing them up for Android Auto.

More importantly, every company in the auto manufacturing and transportation business should soon realize that having an independent provider is the way to go. Why should every company build its own map in a silo, when they could all contribute to a giant crowdsourced map? That's what Open Street Maps (OSM) is for (who would partner today with Encyclopædia Britannica for an information service when you can plug into Wikipedia and contribute to the project)? At some point, public administrations and authorities will probably prefer to work directly with OSM than with proprietary services. Sure, OSM won't be usable for autonomous driving soon but the development of non-proprietary services could have a big influence on the market value of maps (right from the beginning).

 

[Rockster]

If they're planning to make mapping data into a product, it should compel them to build Superchargers faster so that there are fewer mapping voids.

 

[Quant]

Did anyone catch this in the earnings call?

When asked about the value of the data being collected by autopilot, Elon punted by saying that they didn't want to announce new products on earnings calls.

This essentially means that their data is going to be on an order of magnitude better than what anyone else has. So if other car companies want to provide autopilot, that they'd either have to (1) do what Tesla is doing by having all of their cars be beta testers, which is unpalatable for most purchasers, or (2) purchase the mapping data from Tesla.

I believe that this is a new product that will be announced once the data is more refined, in a few months. Thoughts?

Yeah, I caught that and was a bit surprised to read that in between the lines. But, I think Elon is dead set on helping autonomous driving for the world and wants to help the auto industry ....why ? Because TSLA will continue to be 1-1.5 of global volume even 10 years from now ! And he sees safety and saving lives dependent on all vehicles on the road...not just Teslas.

Yeah, once it matures in say 1-2 years, TSLA will sell that as a service....whether we as investors like it or not ! I think Elon believes that only by selling the mapping as a service, will he and TSLA get the millions and millions of cars with the TSLA mapping data feeding back to make its learning engine for mapping way better than any other competitive service.

 

[dakh]

I think selling enhanced maps has been in the cards for quite a while. They have a fairly unique opportunity to gather the data and a data center for storage and analysis. But I suspect there are lots of gaps to be filled before having a marketable product. Firstly, they area mapped is restricted to where Model S cars roam. There are probable millions of miles of roads where they have no driving data. Secondly, they may need some sort of distribution software to push the data out in usable form. So this depends on who the customer is for this data. If they are selling to some GPS navigation service, then not much is needed beyond a data feed. But if Tesla wants to deliver a navigation service to end user, then they will have to do a lot of work. The navigation system currenly in the Model S is in my opinion not good enough to sell to the broad market. For example, Google Map gives much better optionality and monitoring of road conditions. For Tesla to sell a navigation service beyond the captive market of Tesla vehicles, they will need to seriously step up the game and offer something superior to Google Maps and any other GPS provider out there. I seriously hope they do this, because right now I think that navigation and routing are the weakest links in Tesla's AP suite. You seriously cannot have an autonomous vehicle making ridiculous routing choices.

So I am quite enthusiastic about Tesla moving in this direction, but I think there is a lot of work to be done before they have a marketable product.

The value of the data that Tesla collects is tied in with auto-driving. How fast human drivers go around a particular curve for example is fairly irrelevant for a traditional map and is only useful for navigation for maybe estimating arrival time. Yet it's very valuable for AP. It's also something that is not easy to integrate with 3rd party software. But most importantly, why would you give up data that's your strategically critical piece of the puzzle for offering AP? If anything I see Tesla building strategic partnerships and offering the whole package, maybe even including some of the critical hardware pieces.

 

[mdevp]

I heard that question too. But isn't this tech all MobilEye? I never understood what part of AP is Tesla's. I know the chip and system is MobilE and the sensors are Bosch. But what part of AP would Tesla be able to sell? Is the actually data Tesla's or MobilEyes? And how would this data be any different than Google's?

 

[dakh]

I heard that question too. But isn't this tech all MobilEye? I never understood what part of AP is Tesla's. I know the chip and system is MobilE and the sensors are Bosch. But what part of AP would Tesla be able to sell? Is the actually data Tesla's or MobilEyes? And how would this data be any different than Google's?

I used to think that MobilEye, at least in Tesla's case, is responsible for a system that "understands" what cameras see - translates video feed into recognized objects. And Tesla's role is to build up on top of that interpretation of the world a system that would use that to make driving decisions. That would include completely external to MBLY's area inputs like mapping and directions, traffic, possibly knowledge of posted speed limits, etc.

I'm not sure about where that line is anymore, need to study available information a bit more which I don't have the time for at the moment unfortunately.

 

[AlexT]

I heard that question too. But isn't this tech all MobilEye? I never understood what part of AP is Tesla's. I know the chip and system is MobilE and the sensors are Bosch. But what part of AP would Tesla be able to sell? Is the actually data Tesla's or MobilEyes? And how would this data be any different than Google's?

The size of the ever-increasing active Tesla fleet gathering data matters a lot for keeping maps up to date. The current 'Google' based maps aren't that bad, but in my town - outdated (by roughly 2 years).

In addition, I see an opportunity for using Tesla's AP data and "map-gathering" to provide feedback for improving road-signs, markings etc.

 

[replicant]

I used to think that MobilEye, at least in Tesla's case, is responsible for a system that "understands" what cameras see - translates video feed into recognized objects. And Tesla's role is to build up on top of that interpretation of the world a system that would use that to make driving decisions. That would include completely external to MBLY's area inputs like mapping and directions, traffic, possibly knowledge of posted speed limits, etc.

I'm not sure about where that line is anymore, need to study available information a bit more which I don't have the time for at the moment unfortunately.

My understanding: Mobileye learns from the car's immediate surroundings and from the driver's behavior. Tesla acts on this information to communicate with the driver or to drive the car directly - this affects the car behavior, thus helping Mobileye learn. Besides, Tesla takes into account the information provided by cars that already drove through this location.

Mobileye learns from the car's surrounding and driving, whereas Tesla listens to Mobileye and learn from the driving of the entire fleet.

 

[dakh]

The critical piece that I don't yet understand is who is responsible for the interpretation of the recognized data. Say path prediction of various objects, actual steering, braking and acceleration inputs, etc. Based on the fact that from what I can tell Tesla is the only player on the market that can change lanes based on just blinker activation, all of the above is proprietary Tesla implementation. Not to discount what MBLY is doing but that's a pretty big piece.

..also, my understanding is Mobileye is learning how to recognize the stuff that the camera sees, and it might be using the entire fleet of cars it sold for that. They don't necessarily integrated and have GPS signal with that for example. It's a different problem compared to figuring out where to drive and at what speed. That's the part that at least I thought before Tesla has its own implementation for.

 

[dakh]

To put it into an even simpler terms: if a car from the opposite lane is turning left in front of a Tesla, it's MBLY's job to recognize that car and other cars, where the road is etc. And Tesla's job to interpret that as "we're about to run into it based on its current path and our current path, so hit the binders NOW".

What nobody here (that I saw) mentioned yet is I think it's a very ballsy move on Tesla's part to release AP. Say if a car hits emergency brake for no good reason (and car gets rear-ended as a result), or steers itself into something. If it does that fast enough I see it being entirely their fault, nothing a driver could do. That's substantial liability.

Luckily I think a financial filter of being able to afford a Tesla makes it a bit less prone to being piloted by total idiots, plus they seem to have done a good job with the system where it doesn't do anything obviously stupid even in beta version.

 

[Cosmacelf]

Tesla adds high resolution GPS maps onto what Mobileye gives it. The high res maps tells Tesla the center point of each lane on a multi lane freeway or road. This is a huge step up in detail from regular maps and is used in autopilot to integrate with Mobileye lane detection for more accurate lane keeping.

Tesla has a unique advantage over all other car companies in that they gather map data from their fleet of connected cars.

So I suspect the future possible map product is these high res maps, which, as far as I know, aren't available anywhere else.

 

[mdevp]

Cosmacelf (above) posted a really excellent thread on Mobileye/Tesla's AP relationship-
Mobileye and Tesla

 

[replicant]

Could Tesla work with Mapbox for Autopilot 2.0, and ditch Google Maps?

Mapbox Wants To Build Better Maps For Autonomous Cars - Forbes

 

[Spidy]

Isn't MapBox based on OpenStreetMap?

Toyota Selects TeleNav as Brought-In Navigation Supplier with OSM Maps

 

[replicant]

Yes, cf Data Sources on Mapbox Streets | Mapbox