A Step in the Wrong Direction: AP 2.0 Hardware (for now...)

[byan1232]

I just got off the phone with Tesla sales in Paramus, NJ and got some more information regarding the new Model S&X with AP 2.0 Hardware. I was discussing trading in my Model S (pre refresh March 2016) for a 90D Model X with the new hardware. After hearing everything about the new hardware, I will definitely NOT upgrade until at least the new versions are on par with what I have now. This reasons are because:

If you take delivery of the car with the new hardware, you are actually getting LESS features than what cars with AP 1.0 hardware have.

1. Automatic Emergency Braking
2. Collision Warning
3. Lane Holding
4. TACC

Out of these four, to me, TACC and Collision Warning has the most impact. I use TACC on a daily basis (every single time I drive my MS) and Collision Warning, which has proven to be very useful.

With Tesla's track record of over-the-air updates, I doubt the software updates will be on time. This could potentially mean no TACC until early 2017.

Now, the reason why I believe what the sales representative said:

Tesla introduced new hardware to the Model S&X and the company needs to collect new data from this hardware (hence, Elon constantly using the word "shadow") The reason why the 4 features I mentioned are unavailable is simply because they need more data to make sure the initial release is on par with AutoPilot 1.0 hardware cars.

Just a note, I am HUGE Tesla supporter (Model 3 reservation holder) and love driving my Model S. This is just what I think about the product announcement last night.

 

[mrtian97]

good to know. I am hoping to buy Model S when my current lease expires (Aug 2017). I am just hoping they keep the S60 price as current as I don't see myself spending $80k for a car.

 

[Matias]

This is not some secret. Tesla blog

All Tesla Cars Being Produced Now Have Full Self-Driving Hardware

"Before activating the features enabled by the new hardware, we will further calibrate the system using millions of miles of real-world driving to ensure significant improvements to safety and convenience. While this is occurring, Teslas with new hardware will temporarily lack certain features currently available on Teslas with first-generation Autopilot hardware, including some standard safety features such as automatic emergency braking, collision warning, lane holding and active cruise control. "

 

[3Victoria]

So, just wait for the updates to be released, and then upgrade to the new car. No problemo.

 

[Max*]

This is not a step in the wrong direction.

Tesla went from A (AP 1.0) to B (AP 2.0). They took the shorter crowd sourced way, that they usually do, instead of the longer "we'll get 1,000 drivers, give them 1,000 Model S's to drive 1,000,000 miles all over the US and give the general population AP 2.0 in 12 months" route that the typical automaker would take.

If anything, you can argue that their approach getting from point A to point B might not suite your needs, but it's 100% not in the "wrong direction". The wrong direction would go from AP1.0 to nothing.

 

[zer0cool]

I don't see any reason to upgrade until the actual real world functionality is noticeably better. Elon time is generally longer than real time. However I think it's reasonable to believe AP2 to match AP1 functionality within 3 months, since AP1 is basically a subset of AP2 hardware. Functional improvements that I think are next steps would be:
1. Better curve handling
2. Automatically know which lane to be in on the highway
3. Auto merging on the highway to another highway en route to destination

On the highway, merging onto another highway is <0.5% of total travel time. Honestly I don't see it being that big of an improvement. Maybe these changes can come later in 2017. I mean the first AP software took a year to come out. I am not sure why this time would be substantially shorter.

Auto driving on local road is VERY VERY hard. I am not sure I believe this will happen any time soon. Also, even if the tech is there, regulations won't be.

 

[gizmoboy]

I don't see any reason to upgrade until the actual real world functionality is noticeably better. Elon time is generally longer than real time. However I think it's reasonable to believe AP2 to match AP1 functionality within 3 months, since AP1 is basically a subset of AP2 hardware.

Completely agree with this. I would be very unhappy to take delivery of a car tomorrow and not have TACC or emergency braking for months. Given Tesla's track record, December 2016 is not all that likely, though maybe this time they will hit their deadline.

 

[Saghost]

This is not a step in the wrong direction.

Tesla went from A (AP 1.0) to B (AP 2.0). They took the shorter crowd sourced way, that they usually do, instead of the longer "we'll get 1,000 drivers, give them 1,000 Model S's to drive 1,000,000 miles all over the US and give the general population AP 2.0 in 12 months" route that the typical automaker would take.

If anything, you can argue that their approach getting from point A to point B might not suite your needs, but it's 100% not in the "wrong direction". The wrong direction would go from AP1.0 to nothing.

The point he's making is that AP 2 cars will actually be less capable than AP1 cars until Tesla's finishes retraining their neural nets and sends the firmware update out.

Thus a temporary "step in the wrong direction" in terms of capability, however wise it is in the longer term.

 

[Max*]

The point he's making is that AP 2 cars will actually be less capable than AP1 cars until Tesla's finishes retain their neural nets and sends the firmware update out.

Thus a temporary "step in the wrong direction" in terms of capability, however wise it is in the longer term.

I understand the point he's trying to make, I don't agree with it, I thought that was pretty clear with my explanation.

 

[virtualsmack]

1 step back, 2 steps forward. everybody is correct.

 

[bob_p]

It seems likely the missing AP 1.0 features will be enabled fairly quickly (in a few months), since they only use a subset of the new sensors.

 

[number12]

Follow my lead. Wait for a "2017" model car. Say what you will but a Jan 2017 vs a dec 2016 has to command more than just a month premium on the open market.

2017. You get newer year build date. Will be better than Ap1

Thats what I will be doing with my 500 mile old X

 

[HumanGenome]

It's a step in the right direction if the market is willing to pay for it (myself included and probably many more)

You can simply wait until the features are available or purchase an inventory vehicle with AP 1.0 hardware.

If you are in the upgrade boat, just wait until the convenience features are added back.

Am I missing something?

 

[sdorn]

I have seen a lot of people say they are going to wait to order until AP 2.0 is functioning properly, which may not happen until early 2017 (given they are saying December 2016 and are always off on their deadlines). I ordered my car two weeks ago (selecting earliest possible delivery) and my estimated delivery is early 2017, so it seems like you need to order now if you want to get an AP 2.0 car about the time the features come online. Am I missing something?

 

[Joe F]

Don't know when the rest rolls out, but, from the S specs page:

"Safety

• Active safety technologies, including collision avoidance and automatic emergency braking, will become available in December 2016 and roll out through over-the-air software updates."

 

[Dutchie]

If you order now you'll get it he car in December. Right at the time when first features will be implemented. I don't see the issue.

 

[Brando]

good to know. I am hoping to buy Model S when my current lease expires (Aug 2017). I am just hoping they keep the S60 price as current as I don't see myself spending $80k for a car.

Just keep leasing.

 

[S4WRXTTCS]

I'm sure there were arguments at Tesla on how to best migrate to AP 2.0

If you don't release the new hardware till the software is on par you end up continuing to ship hardware where the manufacture of the hardware claims the software exceeds what the hardware was meant to do. It's also risky in terms of the supply chain, and how long you can continue receiving the old hardware. The relationship with the supplier is strained, and every day of using their parts is another day of pain.

If you do release the hardware before the software is ready you're going to be met with a lot of resistance, and if it's late the customers are going to be increasingly livid. It's also a bit risky because in the previous hardware you were using things proven out (TACC, AEB, etc), but with the new hardware it's mostly if not all in-house code. I'm sure they've had people working on for quite awhile, but it's a tough task. If they do ship by Dec 2016 with performance on par with AP 1.0 then all the coders deserve fat bonuses because that would be pretty monumental.

When given those two choices I think it's pretty easy to predict the one Elon Musk picked. It's extremely daring, but at the same time it makes sense to cut-ties with legacy stuff as soon as you can especially if your vision is for so much more than what it can do. Plus if you're Tesla or Apple you can likely get away with it because the user base is so extremely loyal.

My opinion is Tesla made the right call on moving to the Tesla vision hardware so quickly, and I believe the price for the $5K package is correct for what it is.

I'm not in particular agreement with the other $3K software package though because that's stretching things way too much. It could be one of the most important moments in Automotive history though.

 

[ElectricTundra]

It seems likely the missing AP 1.0 features will be enabled fairly quickly (in a few months), since they only use a subset of the new sensors.

Yes, a subset of features on a subset of sensor types, but on an entirely different foundation and likely a different camera. And we don't know how much of that older foundation Tesla has source code level access to and full understanding of.

 

[MP3Mike]

Yes, a subset of features on a subset of sensor types, but on an entirely different foundation and likely a different camera. And we don't know how much of that older foundation Tesla has source code level access to and full understanding of.

My understanding is that even with AP 1.0 the software is almost entirely Tesla written. What wasn't was the camera processing that runs on the EyeQ chip. So I suspect that they have most of it covered already.