Tesla's Autopilot needs to be shut down and NHTSA needs to do their due diligence

[wk057]

I also had AP1 on initial day and specifically recall it being rough and mostly unusable in many situations where it is now fine.

There are plenty of threads and videos documenting this. Consider if you are comparing initial AP2 to current AP1 instead of initial AP1.

I disagree. AP1 on day 1 was far more reliable and predictable than AP2 is months into releases. Day 1 AP1's major problem was diving for exit ramps and turn lanes, a predictable and avoidable behavior. AP2 does all of this plus loses the lanes completely, drifts into oncoming traffic, locks on to the lane and drives into oncoming traffic, dives for barriers and other vehicles, etc.... there's really no comparison. AP1 2 months into releases was infinitely more usable than AP2 is 2 months into releases also. AP1 today, IMO is less usable than previous versions due to the arbitrary speed restrictions, excessive nagging, and purely punitive "features", but that's another discussion entirely.

 

[bigdaddie]

I've got a HW2 car. It works okay on local roads but you MUST exercise extreme caution. It does all the lane deviations mentioned in this thread; however, on well marked roads it works fine. I'm very happy that they do these incremental releases. Bottom line is it is the driver's responsibility to drive the car, even with the diver assist features. It will get there....patience.

 

[Sir Guacamolaf]

.. AP1 on day 1 was far more reliable and predictable than AP2 is months into releases...

Agree. AP2 on week 8 is worse than AP1 was on day 1.

I feel that, AP2,
a) Tesla is starting from almost zero again. The smoke and mirrors marketing however would make think otherwise.
b) They overpromised AP2 big time, it'll never be what they have sold.
c) AP1 is not going to get a whole lotta love going forward. I think Elon tweeted that 8.1 will allow 8.1 to take exits based on nav (when you actually wanted to take exists I mean ) .... yeah March .. still waiting!

 

[Electroman]

AP1 is all I wanted from an auto steer system. To me no more improvments are needed. It does great. I don't care about any other bells and whistles that AP2 may or may not bring.

 

[bob_p]

I hope we'll be getting our AP2 car soon (Tesla evidently has a hold on all 100D deliveries), and we'll be sharing in everyone's frustrations with the limited capabilities (compared to AP1 cars).

Though we should try to put this into perspective - what is the alternative to how Tesla is rolling out the software for AP2?

Tesla could have delayed introducing the AP2 hardware until they had AP2 software working at least as well as AP1. If they had done that, all cars built since October will still be using AP1 hardware. And while the cars would be working better, those owners would miss out on the greater potential of having the AP2 hardware, with the possibility (no matter how remote) that the hardware MIGHT support FSD at some point. And if it takes another 6 months to AP2 working reasonably well - that would be a year's worth of Tesla customers that would have missed out on the new hardware (plus a potential impact on Model 3 production).

Or, Tesla could do what they did with AP1. Instead of trying to slowly enable AP2 functionality, they could wait until they had everything working with at least as much capability as AP1. If they did that, AP2 owners would still have nothing working now that uses the new sensors (which is more than just AP), and we could wait another 6 months or longer before Tesla would have made enough progress with the software to activate the AP2 system.

So, instead of the above, Tesla has tried to roll out AP2 software in phases, with some significant limitations. And, perhaps they missed the target on how they are restricting use of the software.

Ultimately, it is the driver's responsibility to control the car. And if drivers have any concerns at all about the safety of operating under AP, they shouldn't use it. And whenever we eventually get our 100D with EAP/FSD, we'll likely use AP2, but treat it like we are monitoring a student driver, ready to step in and take over, until we have confidence the AP software really knows how to drive safely.

 

[DillyBop]

I'll just chime in with a quick note. I've utilized AP1 since day 1 of public release and have driven tens of thousands of miles with AP1 since that time. It's not perfect, but it's mostly predictable and I generally know where it's going to have trouble.

I recently did ~100 miles with AP2 in an X... And you know what? IMO, this should NOT be publicly utilized nor have even been released yet. Seriously. AP2 was all over the place and completely unpredictable on routes where I've used AP1 without incident over a hundred times. AP2 was nearly unusable. The initial release of AP1 was 20x more reliable, despite AP2 having a significant sensor advantage.

I don't know who in their right mind thought it was a good idea to start rolling out AP2 hardware before they even had basic feature parity with AP1, but those involved should be sacked.

i find it interesting that this post got 3 informative ratings, 2 likes, and 2 loves.

yet it perfectly echoes the OP's sentiment on AP2's shortcomings.

the OP got 32 dislike ratings.

 

[bhzmark]

I disagree. AP1 on day 1 was far more reliable and predictable than AP2 is months into releases. Day 1 AP1's major problem was diving for exit ramps and turn lanes, a predictable and avoidable behavior. AP2 does all of this plus loses the lanes completely, drifts into oncoming traffic, locks on to the lane and drives into oncoming traffic, dives for barriers and other vehicles, etc.... there's really no comparison. AP1 2 months into releases was infinitely more usable than AP2 is 2 months into releases also. AP1 today, IMO is less usable than previous versions due to the arbitrary speed restrictions, excessive nagging, and purely punitive "features", but that's another discussion entirely.

I recall, and it is documented in plenty of videos and complaints on this forum, that early AP1 did ok on straight flat roads with well marked lanes, or in stop and go traffic. But there are loads of complaints that it couldn't take curves or intersections, went lane hunting or drove into ditches or oncoming traffic when lines were missing or shadows on the road or off the highway, had extensive truck lust, etc.

I haven't driven a AP2 so can't directly compare, but I wouldn't fully trust my memory on early AP1 since I've become so used to more fully baked AP1.

But there are still plenty of contemporaneous accounts of the early AP1 and its failings:

First, I don't trust autosteer at all on surface streets, at least for now. It almost ran me into the median after crossing an intersection

some were more charitable back then even AP1 tried to kill, drove into oncoming traffic, drifts out of lane etc.: Contemporaneous notes are more reliable than > 12 mos old recall.

Auto steer only tried to kill us once, but I'll forgive it.

This person wasn't paying attention to the fact that autopilot was already having issues with the glare from what looks like might be a wet road, combined with lots and lots of shadows. You can see it out of his lane at the beginning of the video already and he ignores this.

Eventually auto steer disengages due to lack of confidence in it's data (rightfully so) while telling him to take over. The car then follows the curvature of the road along with the partial left steer that autopilot initiated then canceled, and in this case the car pulls to the left due. His hands aren't on the wheel, so nothing to stop the car from going out of control at this point. User error.

I did test it at low speeds around some sharp curves. It freaks out. Even at 18 MPH and a clear view of the entire curve it won't turn the wheel more than so far while attempting it before "TAKE OVER IMMEDIATELY" happens and it drifts out of the lane. That was a little disappointing. I figured at low speeds it could handle curves. It does, upon "seeing" the curve, immediately pop up with "Hold the wheel", but it will attempt it if you don't, and fail. The turns I tried were pretty easy, but tight. At 18 MPH you don't really get pulled into the door or anything, but I wouldn't take these curves at more than 35 or so, personally.

Well it definitely isn't ready for non beta city driving. They were clear on this. However highway driving works very well in my opinion.

I drove about 10 miles on surface streets with autopilot engaged. A word of warning: This is obviously not recommended by Tesla, and if you want to try this, I *strongly* urge you to be extremely cautious. When I did this, I kept my hands loosely wrapped around the wheel (not touching it, but squeezing my hands would have grabbed it) almost the entire time. There were three times where I had to take control suddenly:

a. The first was when traveling into the sun, and I was on a 4-lane divided country road (2 lanes in each direction). I was in the left lane, and a turn lane split off to the left. The lanes were painted, but the car tried to steer toward the turn lane--it seemed fairly abruptly (more abrupt than I would have turned if I were going into that turn lane). I immediately took over.

b. A little further down the same road--again painted fairly well, but facing the morning sun--the car started drifting to the left too much for my comfort, so I took over. Sun was probably a factor.

c. In the middle of intersections (unpainted lanes), if following a car it seems to do ok, however there is an occasional slightly jerky correction. Without a car to follow, sometimes it worked, and sometimes I felt like I had to take over.

So that's my first experience with autopilot. Overall, TACC is smoother. I had a few pucker factor moments with autopilot, so I urge everyone to use caution for awhile until you really get to know the limits of the system.

Off highway, it often gets confused by intersections, turn lanes, and similar. Sometimes it veers off the wrong way, sometimes it asks you to take over,

Bottom line, it's not really suited for these roads, and was confused by left turn lanes in combination with sweeping corners. In one instance, it jerked over to the right lane, then back to the left lane for no reason. Luckily, I only had one car in front of me. On the plus side, it did slow down for tighter turns, although too late and too abruptly, and not steering smoothly. On the switchbacks on the right, i stopped it as I was not confident it would make the very tight turns.

Having said all that, and knowing it wasn't designed for those kinds of roads,

1) Approaching traffic island (roundabout), it got confused and was going to drive over curb to the right. Fully expected and quick grab back of the wheel required. No warning though. Reinforces the driver is smarter
2) sharp, curvy, road up a hill. Tried repro on this a few times - appears to be worse on sharp right turn going over crest of hill.

 

[Sir Guacamolaf]

@bhzmark as someone who "grew up with AP1" anxiously awaiting every update, measuring each tick and clock and improvement, I can categorically tell you that those two videos were a very extreme example of the first release of AP1.

AP1 on day #1, was better than AP2 is today!

 

[kurdakov]

few videos of AP2 (for non owners of AP2)

 

[S4WRXTTCS]

I'll just chime in with a quick note. I've utilized AP1 since day 1 of public release and have driven tens of thousands of miles with AP1 since that time. It's not perfect, but it's mostly predictable and I generally know where it's going to have trouble.

I recently did ~100 miles with AP2 in an X... And you know what? IMO, this should NOT be publicly utilized nor have even been released yet. Seriously. AP2 was all over the place and completely unpredictable on routes where I've used AP1 without incident over a hundred times. AP2 was nearly unusable. The initial release of AP1 was 20x more reliable, despite AP2 having a significant sensor advantage.

I don't know who in their right mind thought it was a good idea to start rolling out AP2 hardware before they even had basic feature parity with AP1, but those involved should be sacked.

This is an example of someone who knows what they're talking about, and I agree with even the sacking part. What happened was unacceptable.

But, what the OP said?

The OP can't differentiate between AP1 and AP2.
The OP can't seem to understand that we're not at autonomous driving yet. It's still a driving assist package. What he lists are great suggestions for AUTONOMOUS driving, but that isn't what is being shown/developed. Tesla is using the fact that the humans are in charge to justify releasing AP 2.0 with so little testing. Now we could discuss that, but the OP doesn't let us.
The OP seems to dismiss all responsibility of the person behind the wheel. In doing so he ignores how other lane-keeping assist packages operate. That even AP 2.0 isn't much worse than other systems. The MB system for example is dumbed down on purpose.
The OP brings up the McCarth/Speckman crash, but doesn't acknowledge that it was drunk driving. The very kind of thing that has even the NHTSA pushing for autonomous driving.
The OP uses his credentials to give weight to his words, but he doesn't seem to have a clue about driver assist packages, or about the AP system.

The OP comes across as someone wanting to jump on the latest mistake on Tesla's part to try to further whatever objectives he might have. He's so disconnected he can't even get support from someone like me who is extremely critical of how Tesla handled AP 2.0.

I'm sure he's awesome at designing reliable death machines. But, even those machines have had their fair share of problem. It's not like a month or two goes by where I don't here about some issue with some Jet fighter or some other military program having serious issues. Didn't the Aegis cruiser computer system have an divide by zero issue? You'd think of all the issues that could come up that was one they would have caught.

Sunk by Windows NT

There is no doubt that even under the best possible cases that autonomous driving will get some people killed. There is no way around that. The question really comes down to how much better autonomous driving needs to be. We certainly don't need to worry about AP2 killing anyone except out of frustration at how bad it is right now.

 

[S4WRXTTCS]

@bhzmark as someone who "grew up with AP1" anxiously awaiting every update, measuring each tick and clock and improvement, I can categorically tell you that those two videos were a very extreme example of the first release of AP1.

AP1 on day #1, was better than AP2 is today!

But, it was still largely unusable on country roads with curves. I know this because it was one of the first places I tried it on, and the curve wasn't anything significant.

I don't really know what difference it really makes anyways. The problem is when you have confidence in a system there is a greater chance that the confidence is going to result in a crash. Like what happened in Montana when the driver was completely confident in Tesla handling the corner.

If you have next to zero confidence in a system you don't dare trust it. The Montana crash probably wouldn't have happened with AP 2.0

Now I'm not saying this was Tesla's motivation. It certainly has been MB's motivation with their system and they've been clear about it.

I'm simply saying that maybe it's okay that it does suck this badly. This doesn't mean I agree with releasing it, and I certainly don't agree with how TACC brakes for overhead signs. I think that's far more dangerous than the AP 2.0 lane keeping right now.

 

[S4WRXTTCS]

i find it interesting that this post got 3 informative ratings, 2 likes, and 2 loves.

yet it perfectly echoes the OP's sentiment on AP2's shortcomings.

the OP got 32 dislike ratings.

As a general rule you don't bite the hand that feeds you.

WK057 is constantly feeding us. From time to time I do disagree with something he says, but you think I'm going to dislike it? Hell no, and it doesn't matter if it doesn't count for anything.

The only person I'm even less likely to click disagree is ingineer, but that individual tends to be just the facts kind of person. In fact it's so fact based that I have no idea if it's a he or a she. I know there is at least one female Tesla hacker, but I don't know their username.

As to the OP I think he came across very differently. Where he wanted to sink the whole ship instead of focusing on what needed fixing. Now he can argue that the entire ship is infected/bad, but I disagree with that. He also came across as an outsider, and not an owner. I tend to give owners a lot more latitude. I also feel like when I was a non-owner I tended to jump on things where i was critical about something without having experience with it. Once I had experience with it what it may have been wasn't nearly as bad.

 

[kort677]

Hmm.. A 1000 posts doesn't make anyone right and with that assumption, one day members can have very useful comments or thoughts.

The knee jerk part is a different topic. How do you know it's a knee jerk reaction or OP has been contemplating for months, maybe he owns a Tesla with AP?

Bear in mind, i'm not saying he is right or wrong. But shooting down somebody's opinion this way is not constructive in my opinion.

when an opinion is a poor one, coming to poor conclusions it certainly should be "shot" down

 

[calisnow]

Agree. AP2 on week 8 is worse than AP1 was on day 1.

I feel that, AP2,
a) Tesla is starting from almost zero again. The smoke and mirrors marketing however would make think otherwise.
b) They overpromised AP2 big time, it'll never be what they have sold.
c) AP1 is not going to get a whole lotta love going forward. I think Elon tweeted that 8.1 will allow 8.1 to take exits based on nav (when you actually wanted to take exists I mean ) .... yeah March .. still waiting!

a) agree
b) Overpromised on timeframe? Ok. Never be what they have sold? We don't know that yet - at this point AP2 is ahead of where AP1 was at the same timeframe. AP1 had zero self steer functionality 4 months after the hardware release. Personally I'm giving Tesla til October and then I'll judge AP2.

 

[Sully's8]

OP's initial opinion and dare I say "alarmist" post are appreciated. That said, we're (for the most part) rational folks here, and are in part on this forum to monitor the continued progress of the Tesla technology and share our experiences with the vehicle, or in several cases (including my own), learn while we await our vehicle. I would guess that literally nothing in the OP was news to those of us monitoring the progress of AP2 vehicles on Autopilot.

In my opinion the prudence with which Tesla software limits functions and speed are reasonable, and that's the mitigation to the doomsday alarmist point of view.

The active monitoring of the vehicle under the control of AP2 is still required, much like original cruise control functions back in the day, TACC, and other driver assistance features require one to pay attention. Folks, WE WILL ALWAYS NEED TO PAY ATTENTION. Who's kidding who here? The race to vehicles with no steering wheel is underway, but the wheel is there for a reason until further notice.

Similarly, since we don't have a litany of accidents happening (or being being reported), the drivers of the AP2 vehicles are using discretion in engaging AP2.

IMO (eliminating the H), the author of the OP has not oriented his or herself sufficiently to the culture of the forum, the Tesla driving community, nor the Tesla software sufficiently enough to be an authority in the first place. And if that's the case, recommending regulatory intervention is asinine.

So, again, your thoughts are duly noted on this topic and taken with a grain of salt.

Finally, when are you getting your Tesla? Tell us about the features you've chosen and the features to which you're most looking forward. There's a lot more to Tesla obviously than AP2 that most of us here appreciate. What are some of your positives? You are posting on the Tesla Motors Club site, a site for fans and operators of Teslas. You are a driver or a fan, right?

 

[AudubonB]

The OP can thank me for not ripping into the inflammatory and irresponsible title of this post; regardless, it has engendered at least some productive commentary.

The subsequent collective posts also are suggestive that, at one level, the plural of anecdotes can at least sometimes truly be data. Here is mine:

• Yesterday our AP2 Chitty Chitty Shush Shush navigated the 45 miles of a narrow, tight-curved, and moderately steep road.
• It traverses one of the most beautiful stretches of high Sonoran Desert on this planet (otherwise scarcely relevant to this discussion, other than that it provided the reason for performing the rest of the test)
• In that I consider myself fairly familiar with the driving habits of central Arizona's Tesla owners, I can reasonably state it never has been "mapped" (i.e., navigated) by an AP2 vehicle. There is a fairly good likelihood no AP1 car also ever has been on it. It has been navigated, twice, by our AP0 Audiemobile.
• The vehicle's driver was a long-time Mercedes technician who - because long since he became at least as sheltered an Alaskan troglodyte as I - never had even seen any Tesla before, let alone driven in one. I gave him the wheel once we reached this road, and showed him how to engage self-driving
• A safety feature of importance to me (passenger, vehicle owner, reasonably responsible citizen, &c) is that this road is supremely lightly trafficked, other than at mine shift changes. We encountered three vehicles during our drive. I could find a better road than this in Alaska - longer, yet more challenging, lighter traffic - but I'm not there at present so this had to suffice
• Speed limit on the entire length of the road is 35mph, with appropriate yellow "caution" speed reductions at the appropriate curves. Perfect for AP2 at this time.

The result of this test: the vehicle performed perfectly, with the sole exception of one moment when both fog line and center line disappeared and Chitty Chitty hesitated and wandered for a driver-required two-second or so length of time. She hugged the track; she decelerated appropriately at curves and hill crests, there was, other than the aforementioned exception, no time when she was not driving at least as well as any human could drive.


And...my Teutonic friend, who prior to this visit - and this test - had absolutely no intention ever of so doing - ordered a Model 3 last night.

Anyone who wishes to replicate these results is encouraged to traverse the Bagdad Mine Road. Prickly pear blossoms are out and the saguaros will be blooming shortly; this winter's rains have liquified its many stream channels and greened the hillsides to perfection; desert wildflowers are burgeoning daily. The drive provides superb testimony to AP2's effectiveness as of March, 2017.

 

[croman]

I will say, one thing I've noticed is that the "grip" for lack of a better descriptor, on the wheel that AS exhibits on local roads is far looser. In other words, if I pretend like I"m driving the same route as AS, I can affect AS's track by consistently turning the wheel a bit before AS disengages. So I can feather the drive (it drifts too close to one side, I can nudge it back to the center). On highways, AS has a death grip on the wheel so that even the slightest nudge deactivates AS. I think Tesla did this intentionally because local AS is wonky and requires humans to help it along whereas highway AS is much better at staying within lane lines and doing its job.

Just my .02 from observations using AS daily now.

 

[RubberToe]

The OP can thank me for not ripping into the inflammatory and irresponsible title of this post; regardless, it has engendered at least some productive commentary.

The subsequent collective posts also are suggestive that, at one level, the plural of anecdotes can at least sometimes truly be data. Here is mine:

• Yesterday our AP2 Chitty Chitty Shush Shush navigated the 45 miles of a narrow, tight-curved, and moderately steep road.
• It traverses one of the most beautiful stretches of high Sonoran Desert on this planet (otherwise scarcely relevant to this discussion, other than that it provided the reason for performing the rest of the test)
• In that I consider myself fairly familiar with the driving habits of central Arizona's Tesla owners, I can reasonably state it never has been "mapped" (i.e., navigated) by an AP2 vehicle. There is a fairly good likelihood no AP1 car also ever has been on it. It has been navigated, twice, by our AP0 Audiemobile.
• The vehicle's driver was a long-time Mercedes technician who - because long since he became at least as sheltered an Alaskan troglodyte as I - never had even seen any Tesla before, let alone driven in one. I gave him the wheel once we reached this road, and showed him how to engage self-driving
• A safety feature of importance to me (passenger, vehicle owner, reasonably responsible citizen, &c) is that this road is supremely lightly trafficked, other than at mine shift changes. We encountered three vehicles during our drive. I could find a better road than this in Alaska - longer, yet more challenging, lighter traffic - but I'm not there at present so this had to suffice
• Speed limit on the entire length of the road is 35mph, with appropriate yellow "caution" speed reductions at the appropriate curves. Perfect for AP2 at this time.

The result of this test: the vehicle performed perfectly, with the sole exception of one moment when both fog line and center line disappeared and Chitty Chitty hesitated and wandered for a driver-required two-second or so length of time. She hugged the track; she decelerated appropriately at curves and hill crests, there was, other than the aforementioned exception, no time when she was not driving at least as well as any human could drive.


And...my Teutonic friend, who prior to this visit - and this test - had absolutely no intention ever of so doing - ordered a Model 3 last night.

Anyone who wishes to replicate these results is encouraged to traverse the Bagdad Mine Road. Prickly pear blossoms are out and the saguaros will be blooming shortly; this winter's rains have liquified its many stream channels and greened the hillsides to perfection; desert wildflowers are burgeoning daily. The drive provides superb testimony to AP2's effectiveness as of March, 2017.

Audubon,
Sounds like a great drive, did you take a video to share? Do we need to fly Bjorn over to replicate this???

RT

 

[u00mem9]

I hope we'll be getting our AP2 car soon (Tesla evidently has a hold on all 100D deliveries), and we'll be sharing in everyone's frustrations with the limited capabilities (compared to AP1 cars).

Though we should try to put this into perspective - what is the alternative to how Tesla is rolling out the software for AP2?

Tesla could have delayed introducing the AP2 hardware until they had AP2 software working at least as well as AP1. If they had done that, all cars built since October will still be using AP1 hardware. And while the cars would be working better, those owners would miss out on the greater potential of having the AP2 hardware, with the possibility (no matter how remote) that the hardware MIGHT support FSD at some point. And if it takes another 6 months to AP2 working reasonably well - that would be a year's worth of Tesla customers that would have missed out on the new hardware (plus a potential impact on Model 3 production).

Or, Tesla could do what they did with AP1. Instead of trying to slowly enable AP2 functionality, they could wait until they had everything working with at least as much capability as AP1. If they did that, AP2 owners would still have nothing working now that uses the new sensors (which is more than just AP), and we could wait another 6 months or longer before Tesla would have made enough progress with the software to activate the AP2 system.

So, instead of the above, Tesla has tried to roll out AP2 software in phases, with some significant limitations. And, perhaps they missed the target on how they are restricting use of the software.

Ultimately, it is the driver's responsibility to control the car. And if drivers have any concerns at all about the safety of operating under AP, they shouldn't use it. And whenever we eventually get our 100D with EAP/FSD, we'll likely use AP2, but treat it like we are monitoring a student driver, ready to step in and take over, until we have confidence the AP software really knows how to drive safely.

Sir, it appears you may be the last person that still doesn't understand what ACTUALLY transpired here. Tesla had a 2yr supply contract with Mobileye. Leading up to the end of that 2 year period, Tesla and Mobileye could not agree to terms to extend that contract. (Tesla claims Mobileye wanted data, I suspect they just wanted more $ and knew they had Tesla over a barrel...Tesla apparently didn't know how right Mobileye was.) Tesla then successfully marketed a loss of features as a rollout of new technology and started mounting 8 cameras to all new builds...and claiming they will someday allow full autonomous operation.

They are now trying to replicate the function that was lost, which is primarily detecting the road path. Tesla was already responsible for the reaction to that determined path, so replicating the vehicle response that AP1 had is trivial for them. The reason you see AP2 cars left of center and all other types of failure is the vehicle doesn't have a clear understanding of the lane markers. AP1 is vastly superior here, and you can sometimes observe it maintaining autopilot operation even if both painted lane edges are lost.

Realistically, Tesla is not even able to say they have the hardware to support AP1, let alone autonomous FSD. There may be camera / hardware details that Mobileye was using that Tesla failed to comprehend when they attempted to copy and replicate the system in the total thrash to implement something and call it an upgrade.

I think they will eventually get AP2 working at AP1 level. I see no chance FSD is ever deployed. The other thread with the buyer who has requested a refund is one to watch...

 

[Vitold]

Here's AP1 handling road barriers:
(from Tesla Autopilot crash caught on dashcam shows how not to use the system)

AP1 and AP2 are not autonomous and require drivers to think and pay attention.