Upgraded from HW1 to HW2 my thoughts

[chillaban]

I'd say AP1 is like watching a toddler inside the house, and AP2 is like watching the toddler in the front yard.

Both toddlers left unattended could suffer serious injury, one is just more guaranteed that something awful will happen. But in either case, a responsible parent would probably want to keep an eye on the toddler or restrict allowable unsupervised situations.

 

[Marcboy]

Just the thread I wanted to start. My car is in the shop, so I got an AP2 car, and because Tesla is so efficient at sourcing/ making parts, I have been using this AP2 car for a week now, just completed a 300 miles Sunday trip with the family.

In shot, compared to AP1, AP2 sucks.

No auto high beam, no rain sensor. Parking is also not on par with Ap1. I have only seen it work for (sometimes) parallel parking. The one time I used it, it did a great job.

Ap2 refuses to engage in 90% of the non freeway roads, where AP1 works fine. Seems to work better on freeways, but sometimes it does not want to for miles and miles, with clear marking and good weather.
And when it is working, I saw a lot of false positive where the car started braking for no reason, only to re accelerate afterwards

If I had to buy a car now, I would go for AP2 for the promise of a brighter future (I don't think Tesla can even understand the Mobile Eye code, so no chance for any AP1 update), but gawdd this seems like a big step backward.


I read the Kool Aid drinking post about the divorce from Mobile Eye, nice PR work. In the end, it looks like Tesla is taking a very long time to replicate what Mobile Eye had, hence losing on some great new features that would have been added by Mobile Eye in the same time period. We all lose on this.

In the mean time, Ford offers a lot of similar sounding features (I haven't tried them), and some unavailable to Tesla (getting out of parking spot).

Then in 3 months the car is supposed to drive cross country, LOL, not a chance.

I feel bad for anyone who bought FSD, I think that they will get some extra premium features like highway out, but seems like they will never be able to finish their breakfast while the car dries them to work.

Lots of rattles on this new car, but finish seems better, and ultrawhite looks amazing

Last thing, I don't believe one second that regulation is what is keeping Tesla back with AP2. It was approved with AP1, if they could just replicate the behavior, they would get approved.

 

[chillaban]

In the mean time, Ford offers a lot of similar sounding features (I haven't tried them), and some unavailable to Tesla (getting out of parking spot).

If that's how it gets in, I really wanna see how it gets out!

Yes, Tesla is taking a while (if you count a bit over 6 months a while) to replicate what MobileEye has spent nearly a decade perfecting. But the next MobilEye platform isn't getting on the market and available to consumers until 2018+, and AP1 while being great seemed to stagnate in functionality even before the dispute between the companies.

As someone who just recently traded in AP1 for AP2, knowing exactly what I was getting myself into... indeed, being on the ride for the future was a big part of the decision. I consider Teslas a technological purchase, and for those I never like that feeling of being on the peak, no matter how high the peak is.

 

[Marcboy]

For sure AP1 peaked.

Now is Tesla jumping the shark with 8 cameras on board, 1 active ?

I think MobileEye will be easy to bring back in, switch to Intel chipsets, magically they can come with Mobile Eye technology now.

 

[chillaban]

For sure AP1 peaked.

Now is Tesla jumping the shark with 8 cameras on board, 1 active ?

I think MobileEye will be easy to bring back in, switch to Intel chipsets, magically they can come with Mobile Eye technology now.

I mean, believe Elon or not, the claim is that the split was fairly abrupt (ME immediately changed the licensing rules around the EyeQ3 that made it cost prohibitive). But yeah, Tesla basically shipped cars with hardware but zero software ready for AP2.


I don't think ME would be all that easy to bring back in. Intel has been on an acquiring spree of silicon makers. Altera has been a part of Intel for years and none of their products are branded or integrated into Intel's lineup. Intel also bought Infinity the baseband vendor and rebranded their products until the Intel product line, but they contain no Intel IP. I doubt Intel would start asserting control over MobilEye licensing decisions quicker than Tesla can bring AP2 up to parity with AP1 and beyond.

 

[AnxietyRanger]

Last thing, I don't believe one second that regulation is what is keeping Tesla back with AP2. It was approved with AP1, if they could just replicate the behavior, they would get approved.

Obviously it is not regulation, even though Tesla's PR tries to make it sound like that.

If it was regulation holding Tesla back, they could ship a fully capable EAP. Even if it would have the legal limitations of a Level 1 EAP (e.g. hold your hands on the wheel at all times), it would still be able to drive like a Level 5 steeringwheel-less champ.

EAP is bad, because Tesla's AP2 software at this time is bad. They have a separate FSD development "track" that is likely further along in some things, so at some point there will be a step to a next level, but that is nowhere near ready ship yet either.

I agree that the FSD purchase some of us made may well be simply an additional driver's aid for a long time to come, that gets some extra features EAP will be artificially deprived of, but it may not even have any Level 3 (e.g. read a book in limited scenarios) capability for a long, long time.

 

[DavidS314]

Im hoping to get a model s later this year. A big factor in my decision to buy new or used is in wanting AP2. I believe in tesla and believe it will mature in time, but am uncertain in what kind of time line to expect now.

 

[Tam]

...compared to AP1, AP2 sucks...

So far, for the past 1,000 miles, in my experience AP2 is superior in the area of avoiding hitting an obstacle on freeway:

2016 Lebec, CA at 40 mph

2016 Rear ended a van in Switzerland:

There's a current thread citing that:

AP1 Model S crossed yellow line during lane merge


That's should fine as long as AP1 would quit hitting barrier because of ungraceful merging to the right:

In contrast, whenever I see a merging sign on right or left, I have been intentionally activating AP2 auto-lane change to that lane and the system has been merging flawlessly without hitting any barriers:

AP2 has been excellent as shown in the above youtube even with confusing lane markings due to constructions and different road finishes and old scars...

 

[oktane]

My initial impression on AP2 was positive. It was amazing to see that a car can steer and accelerate on its own. It isn't perfect, and is not something I trusted, but a cool novelty. When I got my AP1 loaner, I was blown away by how solid it is. I use and rely on it every day on highways right now.

AP2 -> watching a toddler walk "Wow, way to go little boy! Please be careful"
AP1 -> (not) watching a teenager walk to school "stuff can still happen, but no need to really think or worry about it"

And this is exactly the problem. I was blown away by the AP1 autopilot. When Tesla announced an "enhanced version" of the AP1 autopilot I already thought was useful, and showed a FSD video, I decided to purchase the car.

Total bait and switch to sell us an "enhanced" autopilot and deliver an unsafe autopilot that constantly veers out of the lane and slams on the brakes. They will probably not reach parity with AP1 for 1-2 years (if ever), since the technology behind AP1 took 10 years to develop. By the time they do get it working, every other major car manufacturer will have their own version of autopilot.

 

[AnxietyRanger]

And this is exactly the problem. I was blown away by the AP1 autopilot. When Tesla announced an "enhanced version" of the AP1 autopilot I already thought was useful, and showed a FSD video, I decided to purchase the car.

Total bait and switch to sell us an "enhanced" autopilot and deliver an unsafe autopilot that constantly veers out of the lane and slams on the brakes. They will probably not reach parity with AP1 for 1-2 years (if ever), since the technology behind AP1 took 10 years to develop. By the time they do get it working, every other major car manufacturer will have their own version of autopilot.

I don't it will take 1-2 years for Tesla to reach AP1 parity with AP2, but certainly I agree if it does, that would allow the competition to catch up. Already the Volvo system seems much nicer than AP2 at this time.

 

[oktane]

I don't it will take 1-2 years for Tesla to reach AP1 parity with AP2, but certainly I agree if it does, that would allow the competition to catch up. Already the Volvo system seems much nicer than AP2 at this time.

....as does the Cadillac system with the light bars on the steering wheel and no nag disconnects. I plan on test driving Audi and Caddy this fall. I would never purchase a car without a test-drive, which is why I tested AP1 extensively. My mistake was not knowing AP2 would not build and improve on AP1, but be an in-house disaster.

 

[Snowstorm]

Tesla should really be forth coming on what is exactly going on with AP2. Just tell us the honest truth and let us make the right decision for each of us one way or another.

 

[BrettS]

That's should fine as long as AP1 would quit hitting barrier because of ungraceful merging to the right:

In contrast, whenever I see a merging sign on right or left, I have been intentionally activating AP2 auto-lane change to that lane and the system has been merging flawlessly without hitting any barriers:

AP2 has been excellent as shown in the above youtube even with confusing lane markings due to constructions and different road finishes and old scars...

These were two very different situations. If you look at the road in the first one, both the yellow line on the left and the white dotted line on the right continue straight into the barrier. Certainly autopilot should have noticed the barrier and stopped or attempted to avoid it or something, but autopilot lives and dies by the lines on the road and the lines here guided it straight into the barrier.

On the second video the dotted white line on the right disappears, so the only line it's following is the yellow line on the left, and that line moves over and guides the car into the proper lane. Autopilot doesn't know that it's merging into a new lane. It's simply following the yellow line on the left of the lane. My AP1 car has behaved exactly the same way in the same situation.

This isn't really an AP1 vs AP2 thing, this was a difference in the lines on the road.

 

[electracity]

Tesla should really be forth coming on what is exactly going on with AP2. Just tell us the honest truth and let us make the right decision for each of us one way or another.

What's to tell? They have delivered the software they determined is suitable for customer deployment. They have plans to continuously add features.

Musk says you will be sleeping in the back of your in 2019 if FSD approved by regulators.

What you really want to know, I think, is why the reality of the delivered system seems incongruous with Musk's statements.

 

[AnxietyRanger]

What's to tell? They have delivered the software they determined is suitable for customer deployment. They have plans to continuously add features.

Musk says you will be sleeping in the back of your in 2019 if FSD approved by regulators.

What you really want to know, I think, is why the reality of the delivered system seems incongruous with Musk's statements.

Yeah, the simple, unvarnished truth without hype would be a good start.

 

[electracity]

Yeah, the simple, unvarnished truth without hype would be a good start.

Truth about future events? The developers have lists with hundreds of important items and a plan. There is no truth possible here, only opinion.

 

[Reciprocity]

I had an AP2 for a few days earlier this month before it had to go in to service and I as given an AP1 loaner.
My initial impression on AP2 was positive. It was amazing to see that a car can steer and accelerate on its own. It isn't perfect, and is not something I trusted, but a cool novelty. When I got my AP1 loaner, I was blown away by how solid it is. I use and rely on it every day on highways right now.

AP2 -> watching a toddler walk "Wow, way to go little boy! Please be careful"
AP1 -> (not) watching a teenager walk to school "stuff can still happen, but no need to really think or worry about it"

I really hope parity will come soon as I am pretty spoiled by AP1 now. Please continue with the updates, even if incremental. It gives us hope.

I have not used AP1 more then a test drive. I use AP2 all the time, it is 100% solid on freeways including curves. Only issues I see is when someone is cutting into my lane directly in front of me, AP2 doesn't really acknowledge them until they are directly in front of me.

 

[croman]

I find AP2 lacking in several important areas. Its not worth rehashing but YMMV is pretty much the statement de jure regarding AP2.

 

[Marcboy]

So far, for the past 1,000 miles, in my experience AP2 is superior in the area of avoiding hitting an obstacle on freeway:

AP2 has been excellent as shown in the above youtube even with confusing lane markings due to constructions and different road finishes and old scars...

Not my experience at all. And from seeing the thread, you got the lucky car with the perfect software.

Also, not only that thread, but at least 50% of the latest threads here, on a fan group forum are about AP2, and how disappointing it is.

I was disappointed for sure, others are too whether it is the performance or the missing features or the PR wording.

 

[Reciprocity]

Musk says you will be sleeping in the back of your in 2019 if FSD approved by regulators.

What you really want to know, I think, is why the reality of the delivered system seems incongruous with Musk's statements.

FSD is not AP2. They are not the same animal. Completely different solutions. AP2 is a band-aid because of the souring of the relationship between Mobileye and Tesla after the well known failure of the Mobileye solution to see a giant semi passing directly in front of the car while the driver was not paying any attention to what was going on.

FSD as Elon also said is purely camera/GPS based solution. One of the downsides of having to many sensors and trying to rely on them all, is which sensor is right? When they are all inadequate, which one do you rely on? I believe this is why Tesla is going with a Vision only solution. GPS is used to find our general location, then the Vision system identifies landmarks to identify where you are on a high-def map and to get to the kind of margin of error you need to drive.