Autopilot 2.0, what do we know about it?

[krazineurons]

I am considering to buy a MX or a MS this month and i am wondering if my investment to buy a Tesla right now would mean i would be driving a soon to be outdated vehicle in less than 2 years.

I have been following the M3 (Model Y) news and it appears that it might be Level 4 Autonomous capable (Autopilot 2.0). It appears that there would definitely be new hardware to support Autopilot 2.0, but what is not clear if existing cars would get Retro fit with the new hardware.

What does the tesla brain-trust think about it?

 

[dandurston]

I think Autopilot 2.0 is coming some time late this year or next, but it won't be Level 4. Tesla has long talked about the gradual road to full autonomy, with lane keeping at one extreme (highway) and summon at the other (urban parking) of the spectrum with numerous points in between.

I think Autopilot 2.0 will have more forward facing cameras and a proper rear facing camera so it can do stuff like change lanes without you having to manual verify it's clear (right now a rapidly overtaking car is a danger). This will allow it to change lanes at will on the highway so it can hold a cruising speed with no involvement from the driver.

Tesla says 2 years for full automony but this is an area that is notorious for problematic 1 in a million incidents that are hard to resolve. I think we'll see a steady march towards it with full autonomy around 2019-20.

 

[Saghost]

Well, there's this:

New Tesla Model S Has 2nd Triple Cam For Autopilot & Pedestrian Noise Unit

It seems fairly likely that an improved Autopilot hardware set will show up either this year or next.

Another random tidbit: the X camera installation isn't identical to the S one (and seems to have a space for a second forward facing camera of similar size,) but the S version did not get changed during the refresh to match the X type.

My guess/hope is that if they do release a new module this year it'll be upgradeable on at least the X. If it doesn't come out until next year, anything goes.

 

[Alketi]

If autopilot is important to you, then I would wait ~6 months. Elon said last week there will be a "major" announcement at the end of the year and retrofitted Tesla's are already appearing in the wild.

 

[calisnow]

I am considering to buy a MX or a MS this month and i am wondering if my investment to buy a Tesla right now would mean i would be driving a soon to be outdated vehicle in less than 2 years.

Extensive evidence points to an updated hardware suite before the end of 2016. This most likely will consist of at the very least several additional cameras - and likely more radar units as well.

I have autopilot now and it is excellent - the problem is that it is SO good you will wish it had more cameras to do things like cross intersections, see further ahead on the freeway etc. - If I were you and I wanted to keep my car for 5 years or more I would wait another 6 months. You will be glad you did.

 

[ecarfan]

I am considering to buy a MX or a MS this month and i am wondering if my investment to buy a Tesla right now would mean i would be driving a soon to be outdated vehicle in less than 2 years.

Tesla improves its car continuously and does not announce changes in advance. Whenever you buy a Tesla you can be sure that in 2 years the cars will be much better. So you can wait forever, or buy now.

No one knows when full autonomous driving capability will be available. Not even Tesla at this point.

 

[calisnow]

Tesla improves its car continuously and does not announce changes in advance. Whenever you buy a Tesla you can be sure that in 2 years the cars will be much better. So you can wait forever, or buy now.

No one knows when full autonomous driving capability will be available. Not even Tesla at this point.

Sorry this simply is not true at this point and you know it. Don't mislead the newbies with that generic advice when there is better intel available.

1 - Elon has said there will be the big "reveal part 2" at the end of 2016 for Model 3. 99% that is going to be about autonomous capabilities. Elon will not tank S&X sales in December by telling the world that the $35K Model 3 will have far greater capabilities - the only way this announcement works is if the S&X have autonomous hardware upgrades at the time of the Model 3 reveal part 2.
2 - JB Straubel recently stated the S & X will lead the way with tech vs the 3
3 - The 3 is due theoretically in July 2017
4 - Leaked schematics of recent S's show wiring for a three camera set-up.
5 - The current autopilot hardware will be two years old in October
6 - Mobileye's execs have gone on the record in multiple public presentations over the last 12 months stating that one OEM is coming out with an 8 camera, 5 EyeQ3 SOC based system in 2016 or 2017.
7 - Mobileye's execs have stated that the 8 camera system, combined with radar and sonar, will be all that is needed, hardware-wise, for full autonomy - and that it will be software development from there.

Put 1-6 together and we have, in all likelihood, the making of a major autopilot revision coming in 8 months or less that will likely carry the cars forward for several years.

This is in fact an unusually bad time to buy an S or an X if you can wait.

 

[M0DEL³]

Sorry this simply is not true at this point and you know it. Don't mislead the newbies with that generic advice when there is better intel available.

1 - Elon has said there will be the big "reveal part 2" at the end of 2016 for Model 3. 99% that is going to be about autonomous capabilities. Elon will not tank S&X sales in December by telling the world that the $35K Model 3 will have far greater capabilities - the only way this announcement works is if the S&X have autonomous hardware upgrades at the time of the Model 3 reveal part 2.
2 - JB Straubel recently stated the S & X will lead the way with tech vs the 3
3 - The 3 is due theoretically in July 2017
4 - Leaked schematics of recent S's show wiring for a three camera set-up.
5 - The current autopilot hardware will be two years old in October
6 - Mobileye's execs have gone on the record in multiple public presentations over the last 12 months stating that one OEM is coming out with an 8 camera, 5 EyeQ3 SOC based system in 2016 or 2017.
7 - Mobileye's execs have stated that the 8 camera system, combined with radar and sonar, will be all that is needed, hardware-wise, for full autonomy - and that it will be software development from there.

Put 1-6 together and we have, in all likelihood, the making of a major autopilot revision coming in 8 months or less that will likely carry the cars forward for several years.

This is in fact an unusually bad time to buy an S or an X if you can wait.

Speaking of "better intel": The 3 is not due in July 2017. It is due at the end of 2017, the July 2017 date is the date suppliers are expected to have parts delivered to Fremont by.

The OP's question was not about "Autopilot revisions", but rather Full (Level-4) autonomous operation. So most of what you offered above is either non-sequitur or conjecture to the question at hand. At this point there is no assurances from Tesla that AutoPilot 2.0 hardware will eventually support Level-4 Autonomy. (I personally do not believe it will.) And even if it does, there is no time-line published as to when Tesla will activate it in software. Finally, there is a regulatory process that will have to occur, after everything works, before it will be available to the public.

ECarfan's response of "No one knows when full autonomous driving capability will be available. Not even Tesla at this point." remains the best response to the OP's actual question.

 

[calisnow]

What I posted are a series of signs that autopilot 2.0 (at least the hardware) is almost upon us (ie 8 months or less) - those of us who buy Model S & X anyway.

All signs point to an imminent AP hardware change dude - and the OP asked about capabilities - not operation. If you were planning to spend $100K on a Model S and were concerned about hardware becoming obsolete you would care very much about the kind of signs which show that the autopilot hardware is about to get revamped.

And what you are is some guy typing on the internet (as is the other poster on this thread who claims Tesla doesn't know ) arguing about level numbers - when in fact the two public data points from people whose opinions actually matter are:

1 - Mobileye has stated that the 8 camera system on the way through one OEM (likely Tesla) will be hardware capable of full autonomy - and that it's only software development after that (look up the quote yourself).

2 - Musk has publicly stated the cars will be capable in two years - if he follows past practice the hardware will come first, followed by a long testing/training period. That time frame dovetails with an imminent change to the autopilot hardware followed by a 12-18 month fleet testing period before the features go live and fits his "two year" time frame prediction.

I have one Model S with AP already, and I have put an order for a second one on hold because of the high probability that hardware version 2.0 is just around the corner - other owners are doing the same.

If the OP is planning on keeping his car for a long time he is better off waiting a few more months.

Note that all the keyboard jockeys who are so concerned with what the regulators say fail to note that autopilot 1.0 showed up and got turned on before regulators had anything to say. A more capable version with more cameras will likely get turned on and simply be another "beta" but in reality will be full autonomy - just without legal approval for drivers to actually close their eyes - but many will do so anyway regardless of whether or not the law has caught up.

Many of us current owners do things like check e-mail etc. on autopilot 1.0 - even Elon has joked about doing it himself.

 

[ItsNotAboutTheMoney]

Very little. There's speculation that a picture of an S with extra camera gear in front is related to it.

That's it.

 

[Saghost]

The problem with this clear case you're building is that the details are fuzzy and don't align...

The schematic was very specifically for a triple camera in front, no additional radar or other cameras. That doesn't mesh well with the eight camera system quote.

Elon said another reveal closer to production, maybe at the end of the year - which means it might be sometime next year, too.

There's no question better hardware is coming. I think there's a lot of question about the six months part...

 

[AZ Desert Driver]

The problem with this clear case you're building is that the details are fuzzy and don't align...

The schematic was very specifically for a triple camera in front, no additional radar or other cameras. That doesn't mesh well with the eight camera system quote.

Elon said another reveal closer to production, maybe at the end of the year - which means it might be sometime next year, too.

There's no question better hardware is coming. I think there's a lot of question about the six months part...

==Sorry I'm so clumsy about linking posts↑====[from a parallel thread- I asked===↑
Same question - stated differently....I want my MS NOW....but I also know that V2 is close. V2 may actually be undergoing stealth deployment now - as in having the wiring harness factory built to accommodate everything in use now and fairly certain use in the near future. What I would be willing to do - is to buy the prototype cameras and wiring NOW, even if the V2 switch is not flipped for many more moons.

I am ordering the new MS60 now - to take advantage of the Referral, 60/75 kWh , $7500 tax incentives. What I want to do is delay production until Sept (this is my anniversary present) and ...perhaps into Oct if there is any substance to these V2 rumors.

 

[calisnow]

==Sorry I'm so clumsy about linking posts↑====[from a parallel thread- I asked===↑
Same question - stated differently....I want my MS NOW....but I also know that V2 is close. V2 may actually be undergoing stealth deployment now - as in having the wiring harness factory built to accommodate everything in use now and fairly certain use in the near future. What I would be willing to do - is to buy the prototype cameras and wiring NOW, even if the V2 switch is not flipped for many more moons.

I am ordering the new MS60 now - to take advantage of the Referral, 60/75 kWh , $7500 tax incentives. What I want to do is delay production until Sept (this is my anniversary present) and ...perhaps into Oct if there is any substance to these V2 rumors.

If you switch cars every couple years I guess it doesn't matter much when you get your Model S - but if this is a long term purchase you plan to keep for 5-10 years or more then in your shoes I would delay production until the big announcement that Elon has publicly stated is coming at the end of 2016. That will be Model 3's "part ii" unveiling - and most likely will involve autonomy details for Model 3. At that time Model S/X will most likely get hardware upgrades for the next generation of autopilot. So, by waiting another 90 days you will probably get at the very least a tri-focal camera for the front which will significantly enhance your autopilot's long term potential capabilities even if the software is not ready to go in December.

 

[ohmman]

While I think @calisnow's conclusion could play out, I don't think it's anywhere near the confidence level where I'd change my timing on a vehicle purchase. Model 3 won't be built and sold until 2017 (to employees) and sometime in 2018 for the general public. If it's anything like the X rollout, there will be token 2017 deliveries and a slow ramp. They don't have to release AP 2.0 in the S and X until that time, assuming that part 2 truly shows a tremendous gain in that particular tech. So assuming that we see a big improvement in the part 2 reveal, and that's an assumption based solely on the words "the obvious thing," S and X sales will continue with the current suite because they can just say "this hardware will be available in the Model 3s that roll off the line."

I think @ecarfan's response wasn't generic, it was the more responsible of the two. Connecting some widely scattered dots to draw a picture is a fun and valuable exercise, but to imply others should make financial and life decisions based on those is probably not the best advice.

 

[Cosmacelf]

My personal opinion is that AP V2 hardware won't be in new cars until end of 2016 at the earliest, and could easily be later. Bear in mind that manufacturers fix hardware well in advance of the software being done at their own peril. If I were Tesla, I wouldn't fix the hardware (meaning, allow no more changes to it) until the software had gotten to a good point, since the software engineers could hit a roadblock, which could be fixed with a small hardware tweak.

For instance, it is very possible that full autonomy will require more CPU power than today's best engineering estimates. If hardware is fixed in place now, and down the road, you run out of CPU speed to actually do full autonomy, that's going to suck.

 

[aesculus]

I wouldn't fix the hardware (meaning, allow no more changes to it) until the software had gotten to a good point, since the software engineers could hit a roadblock, which could be fixed with a small hardware tweak.

When has that ever happened?

 

[ecarfan]

No one knows when full autonomous driving capability will be available. Not even Tesla at this point.

Just to clarify my thinking on this: while there is a great deal of speculation about "AP V2" coming out "soon", as in maybe by the end of this year, it is even more speculative to assume that "AP V2 = full autonomous driving".

I suspect it will not be full autonomous, as in ", with the car parked, someone sitting in driver's seat enters their destination into the car nav and the car then drives to that destination over many miles over all kinds of streets and highways without the person in the driver's seat touching the controls".

I think it is far more likely that AP V2 hardware -- and all future firmware upgrades that use that hardware -- will significantly extend the AP capabilities Tesla offers right now, but that hardware will be insufficient to allow full autonomous driving as I describe it above. That will require an AP V3 hardware that is at least a few years away.

Of course, I could be wrong...

 

[Alketi]

Personally, unless pressed by external constraints, I would never buy hardware before an imminent inflection point.

The "always improving" argument doesn't hold water because there are occasional inflection points where capability leaps beyond the usual incremental improvements. AP V2 will be one of those inflection points.

Musk stated to Walt Mossberg that his end-of-year announcement would be "real big news" if he gave it now, in direct reference to whether the Model 3 would support autonomous driving. He's also stated that the S/X being the technology-leaders will always get the latest technology first. Thus, the conclusions seem fairly obvious to me.

To put it another way -- the penalty for waiting and not getting AP V2 is just ~6 months of your time.

The penalty for buying now and then seeing AP V2 hardware delivered end-of-year is resale-loss + hassle + time if you decide to sell/swap your car, or flat-out not having the feature for years if you decide to keep your car long term.

 

[ohmman]

To put it another way -- the penalty for waiting and not getting AP V2 is just ~6 months of your time.

The penalty for buying now and then seeing AP V2 hardware delivered end-of-year is resale-loss + hassle + time if you decide to sell/swap your car, or flat-out not having the feature for years if you decide to keep your car long term.

A hypothetical announcement of AP V2 doesn't necessarily mean hardware being delivered contemporaneously. It could just be "we've come far, and by the time the 3 is offered, we'll have this feature set." Keeping in mind those could be Tesla timetable.

So your options aren't inclusive of the wide variety of true outcomes. The penalty for waiting and not getting AP V2 might be that you're that much closer to AP V2. And the rumor mill will be churning much more aggressively. So it could reset the timer and you wait another 6 months, etc. I'm not suggesting this is what will happen, but it's clearly in the realm of possibilities. So it's not just a handful of months of your time, it's an indeterminable period of time that you're without the car.

There's also an argument to be made that whatever leap AP V2 brings, it's still only going to be iterative en route to a driverless car. It's not going to be the driverless car. Currently, AP functions remarkably well for a lot of people. I don't live in a big congested city, where the next iteration might be very welcome. Driving around my smallish town, I have little (if any) desire to use autonomy. Highways are so miserable that having AP on them covers about 95% of what I'd want anyway. I can't imagine I'm alone.

So the way I see it, yes - there is some risk. But I don't see it as being either as imminent as you do, nor as much of an inflection point. We'll have to disagree on that one.

 

[DrivingTheFuture]

I think it's safe to say we know it will be freaking awesome. However after personally using the current version of AP for several hours a few weeks ago on a roadtrip, I now think of AP and machine learning quite often while I'm driving. I must say if the capability to handle complex intersections becomes available, I think I will be much more wary of trusting it initially like I did on the highway. There are SO many variables... but I trust Elon's comments on the matter and am sure they are based off the incredible things he has seen / witnessed in meetings with MobileEye.