.48 feels like AP2 finally passed AP1

[AnxietyRanger]

can we add a few more?:

- AP2 does not have multi-type vehicle ID display, where there is no passenger in the front seat
- AP2 does not have multi-type vehicle ID display, where there is passenger in the front seat
- AP2 does not have multi-type vehicle ID display, in day time
- AP2 does not have multi-type vehicle ID display, during night time
- AP2 does not have multi-type vehicle ID display, when a man is driving
- AP2 does not have multi-type vehicle ID display, when a woman is driving

You can't refute that in all of these scenarios AP1 does have multi-type vehicle ID display.

Oh my, the list is now a dozen long. I betcha if you dig deeper you can get a few more.

You may have missed that three of my bullets were not "multi-type", but two were about "multi-lane".

They were all distinctly different features.

I know it bugs you that AP2 has not actually reached AP1 parity. Get over it and embrace the facts.

AP2 will get there and beyond one day. That day is not just here yet.

 

[chillaban]

I think it would be great if they got autopilot to avoid driving in adjacent vehicle's blind spots. Is that coming some day?

It would be nice. I completely agree though — the biggest class of disengagements for me, on AP1 and AP2, is when I enter a sharp curve with a truck or erratic vehicle to my side. And when I'm piloting the car, I would never attempt to drive beside a car like that.

Some amount of gap spacing would be nice, at least until AP1/AP2 can demonstrate that their side collision mitigation actually works reliably enough to depend on it.

 

[TaoJones]

That lane change issue must be situation dependent. I've done hundreds of lane changes on non-freeway 2 lane highways. As of .42 I found AP2 to be comparable / superior to AP1 on the same highway.

Divided highways? Could these be state routes?

I haven’t had a single AP2 lane change on non-divided roadways even though there are more lanes on those roads than most state routes or of course highways.

Tangentially, am starting to think it might be worthwhile to cobble together a short list of proving ground routes - e.g., the shrimpster’s Route 2222 in Austin, Portuguese Bend on the outer edge of the Palos Verdes Peninsula here in LA County/South Bay, any 10 miles of LA’s finest poorly-marked freeways (I vote for that extra-special section of the 605 adjacent East LA that arbitrarily drops to 45mph), and the list goes on.

Geocaching for Teslas, as it were. Tokens can be left at nearby SCs similar to when that fellow used to leave coins or some such a couple of years ago.

 

[Joe F]

It is a quest for facts and accuracy. As well as a visual representation of the identification abilities of the products.

If we're going to say AP2 has AP1 parity or exceeds it, we better get the facts right.

Others can decide how much they value each of those differences.

So to you, pretty pictures makes for equality. Okay.

Who are you to decide the features between AP1 and AP2 should behave in the exact same way, despite it meaning nothing to the operation of the vehicle? Somewhat petty, IMHO. Methinks you’re running out of things to whine about. What’s next, font differences?

If you can’t see beyond the windshield, and have to fixate on the IC to determine if the blob in front of you is another car, or truck, or perhaps a motorcycle, may I respectfully suggest you need to have your vision checked.

 

[Electroman]

It is a quest for facts and accuracy.o

It's a quest for whining for whining sake.

 

[EV-lutioin]

I guess so!

It's ok to turn your phone sideways... won't hurt it a bit.

 

[TaoJones]

So to you, pretty pictures makes for equality. Okay.

Who are you to decide the features between AP1 and AP2 should behave in the exact same way, despite it meaning nothing to the operation of the vehicle? Somewhat petty, IMHO. Methinks you’re running out of things to whine about. What’s next, font differences?

If you can’t see beyond the windshield, and have to fixate on the IC to determine if the blob in front of you is another car, or truck, or perhaps a motorcycle, may I respectfully suggest you need to have your vision checked.

Attacking the messenger because you don’t like the message doesn’t change the message.

Neither does attempting to diminish the functionality, nor does asserting that it’s okay to get less with AP2 than we had with AP1.

Fact is that over a year after we were told AP2 would be at parity, and over a year after we were shown a now-debunked video of what AP2 was going to do for us, AP2 still isn’t at parity with AP1. Even the most ardent apologists and fanbois have had to concede, in deference to those who actually know what they’re talking about (see several posters In this thread who a) actually own Teslas, b) have owned AP1 and AP2 cars, and c) drive more than 100 miles a week and who might even leave their own county once in awhile for something other than a wedding or a funeral.) that the jury is still out.

I’ve identified one use case in which AP2 was superior to AP1 as early on as this past May, only 6 months into its public cycle. Others have noted that at least on the highway under admittedly vanilla conditions, they can’t tell the difference between AP1 and AP2.

One of the reasons why IC differentiation matters is because Tesla is behind when it comes to reacting to small humans and basketballs that might cross the path of our chariots - ostensibly at low speeds.

With AP1, you knew that MobilEye could tell the difference between a semi, a car, and a motorcycle. We call that progress toward the objective of not running over granny when she trips over her walker and takes a header into the street in front of us.

With AP2, we don’t know. Actually, we do know, and it doesn’t (see the work of verygreen and others who’ve rooted their chariots).

So it’s not just about kvetching to kvetch. It’s about being sold/told more and receiving much less than what we had. Now it’s a race to deliver *something* that resembles past promises before the lawsuits increase (class actions have already begun).

Personally, I think AP4 will be stellar in 3 years. So does Elon who pushed his FSD estimate to 2020. I also think we’ll get something decent by mid-2018 for AP2/EAP. As a fallback, I’m actively looking for an AP1 car to replace my AP2 mistake because the next 6 months will offer some great low-mileage lease returns as CPOs. So either way, in 6 months, things will be better.

Meanwhile, AP2 remains, in a word, brittle.

 

[TaoJones]

It's a quest for whining for whining sake.

Polly, meet Anna.

Don’t forget to bring enough Kool Aid for everybody.

 

[EinSV]

Attacking the messenger because you don’t like the message doesn’t change the message.

Polly, meet Anna.

Don’t forget to bring enough Kool Aid for everybody.

Pot, kettle.

 

[TaoJones]

Pot, kettle.

Helpful as always, I see.

Ass, kiss mine. Do try to differentiate between respective posts and replies.

 

[jimmy_d]

Divided highways? Could these be state routes?

I haven’t had a single AP2 lane change on non-divided roadways even though there are more lanes on those roads than most state routes or of course highways.

Tangentially, am starting to think it might be worthwhile to cobble together a short list of proving ground routes - e.g., the shrimpster’s Route 2222 in Austin, Portuguese Bend on the outer edge of the Palos Verdes Peninsula here in LA County/South Bay, any 10 miles of LA’s finest poorly-marked freeways (I vote for that extra-special section of the 605 adjacent East LA that arbitrarily drops to 45mph), and the list goes on.

Geocaching for Teslas, as it were. Tokens can be left at nearby SCs similar to when that fellow used to leave coins or some such a couple of years ago.

Oh did I misunderstand? I assumed the statement was about divided highways. A lane change on a 2 lane non-divided highway puts you into oncoming traffic, doesn't it? That seems like something they probably don't want right now. I don't specifically recall if I've tried lane changes on non-divided highways, but I did a 7000 mile road trip on AP1 through the western US last summer and I don't recall lane changes not working anyplace that I expected them to. There was a lot of undivided highway on that trip.

My benchmark route is downtown LA to downtown SF and back (every couple of weeks for maybe 20 years now). It includes a lot of different kinds of driving, but it's all divided highways. I've driven it hundreds of times without AP and dozens of times with AP. On AP1 I knew all the failure points by heart, but AP2 has been changing fast since I got it - every time fewer failures. 42 still had more failures on AP2 than on AP1 (which has about a half dozen places where it's likely to require an intervention). 42 on AP2 requires about 15 interventions over 6 hours / 400 miles - over half of them are on the grapevine through the mountains and most of the rest are on twisty bits of freeways in the East Bay. I haven't had the courage to drive AP through the tollgates on the bay bridge yet but I expect it could do it at the posted 25MPH limit. If 48 can get me down to single digits then it'll be in the same ballpark as AP1 - for me.

 

[TaoJones]

Oh did I misunderstand? I assumed the statement was about divided highways. A lane change on a 2 lane non-divided highway puts you into oncoming traffic, doesn't it? That seems like something they probably don't want right now. I don't specifically recall if I've tried lane changes on non-divided highways, but I did a 7000 mile road trip on AP1 through the western US last summer and I don't recall lane changes not working anyplace that I expected them to. There was a lot of undivided highway on that trip.

My benchmark route is downtown LA to downtown SF and back (every couple of weeks for maybe 20 years now). It includes a lot of different kinds of driving, but it's all divided highways. I've driven it hundreds of times without AP and dozens of times with AP. On AP1 I knew all the failure points by heart, but AP2 has been changing fast since I got it - every time fewer failures. 42 still had more failures on AP2 than on AP1 (which has about a half dozen places where it's likely to require an intervention). 42 on AP2 requires about 15 interventions over 6 hours / 400 miles - over half of them are on the grapevine through the mountains and most of the rest are on twisty bits of freeways in the East Bay. I haven't had the courage to drive AP through the tollgates on the bay bridge yet but I expect it could do it at the posted 25MPH limit. If 48 can get me down to single digits then it'll be in the same ballpark as AP1 - for me.

Ok, hold on here - since you’re local , do you mean divided highways to include only the 405, 110, 101, or to also include, say, Hawthorne or Sepulveda or some other boulevard with both medians and turn lanes as dividers (trying to come up with streets other than 2-lane meaning 1 lane in each direction - which neither of us meant).

I’ve been able to auto lane change fine heading up to SF via the 5 or 101, but not on the wider parts of PCH/Route 1. This is the same as what you’ve experienced, yes?

Also with regard to your previous question, AS especially with Chill on *is* much better with .48 compared to .42 relative to lane wandering. Problem is it still crosses lane markers and gets too close to adjacent traffic. Optimistically, once the other cameras are brought into play, the car should track pretty much perfectly.

 

[jimmy_d]

Attacking the messenger because you don’t like the message doesn’t change the message.

Neither does attempting to diminish the functionality, nor does asserting that it’s okay to get less with AP2 than we had with AP1.

Fact is that over a year after we were told AP2 would be at parity, and over a year after we were shown a now-debunked video of what AP2 was going to do for us, AP2 still isn’t at parity with AP1. Even the most ardent apologists and fanbois have had to concede, in deference to those who actually know what they’re talking about (see several posters In this thread who a) actually own Teslas, b) have owned AP1 and AP2 cars, and c) drive more than 100 miles a week and who might even leave their own county once in awhile for something other than a wedding or a funeral.) that the jury is still out.

I’ve identified one use case in which AP2 was superior to AP1 as early on as this past May, only 6 months into its public cycle. Others have noted that at least on the highway under admittedly vanilla conditions, they can’t tell the difference between AP1 and AP2.

One of the reasons why IC differentiation matters is because Tesla is behind when it comes to reacting to small humans and basketballs that might cross the path of our chariots - ostensibly at low speeds.

With AP1, you knew that MobilEye could tell the difference between a semi, a car, and a motorcycle. We call that progress toward the objective of not running over granny when she trips over her walker and takes a header into the street in front of us.

With AP2, we don’t know. Actually, we do know, and it doesn’t (see the work of verygreen and others who’ve rooted their chariots).

So it’s not just about kvetching to kvetch. It’s about being sold/told more and receiving much less than what we had. Now it’s a race to deliver *something* that resembles past promises before the lawsuits increase (class actions have already begun).

Personally, I think AP4 will be stellar in 3 years. So does Elon who pushed his FSD estimate to 2020. I also think we’ll get something decent by mid-2018 for AP2/EAP. As a fallback, I’m actively looking for an AP1 car to replace my AP2 mistake because the next 6 months will offer some great low-mileage lease returns as CPOs. So either way, in 6 months, things will be better.

Meanwhile, AP2 remains, in a word, brittle.

I have a somewhat different perspective, but I don't really disagree with you. I started this thread because I wanted to hear what kinds of experiences people were having with recent AP2. I have my experiences, and my needs, and they lead to my impressions. Other drivers have different circumstances and experiences.

I think we all have a lot in common, but of course differences are what show up in debate. Some of us feel justifiably betrayed by stuff Tesla has failed to deliver on, and others of us are just delighted to have a really great car.

 

[jimmy_d]

Ok, hold on here - since you’re local , do you mean divided highways to include only the 405, 110, 101, or to also include, say, Hawthorne or Sepulveda or some other boulevard with both medians and turn lanes as dividers (trying to come up with streets other than 2-lane meaning 1 lane in each direction - which neither of us meant).

I’ve been able to auto lane change fine heading up to SF via the 5 or 101, but not on the wider parts of PCH/Route 1. This is the same as what you’ve experienced, yes?

Also with regard to your previous question, AS especially with Chill on *is* much better with .48 compared to .42 relative to lane wandering. Problem is it still crosses lane markers and gets too close to adjacent traffic. Optimistically, once the other cameras are brought into play, the car should track pretty much perfectly.

I'll definitely give Chill a try on my next run.

I don't consider Hawthorne or Sepulveda to be highways, I consider them to be surface streets. That's just my own labeling though. Of course there's a whole continuum of stuff here in LA county but personally I wouldn't consider something to be a proper highway unless it has fairly limited access, few controls (lights, stop signs), a speed limit of 55 or higher, and few intersections in general. Some roads go back and forth depending on where you are on them. As a rule I don't expect AS to work in the vicinity of intersections, which might be influencing my sense of what constitutes a highway. I use TACC on surface streets all the time, but I don't trust AS on surface streets without really close supervision.

When I think of a multi-lane undivided highway I'm thinking of something like the parts of the 1 where there are 2 lanes going each way, but where you're miles away from the nearest intersection - or maybe the parts of the 39 or the 2 through the mountains.

Now that I think of it, it's hard to come up with good examples in LA.

 

[Joe F]

Attacking the messenger because you don’t like the message doesn’t change the message.

[snip]

With AP1, you knew that MobilEye could tell the difference between a semi, a car, and a motorcycle. We call that progress toward the objective of not running over granny when she trips over her walker and takes a header into the street in front of us.

The message is flawed.

So how, exactly, is granny depicted on your AP1 IC? Better yet, what, exactly does your AP1 car do with the pretty pictures it displays on your IC? If there’s no operational decision made based on a (sometimes incorrect, I’ve read) detected vehicle type, of what utility is it?

Maybe that’s why we have that nice, big screen with a drawing program; To appease those who like pretty pictures. Apparently it’s important to some.

 

[EinSV]

Helpful as always, I see.

Ass, kiss mine. Do try to differentiate between respective posts and replies.

Doubling down on polemics and personal attacks, nice.

Given all the AP2 videos floating around that make you and the rest of the AP2 hater-brigade look ridiculous, I can understand why.

Perhaps you can stick with facts and, better yet, videos. How about some AP1 and AP2 video comparisons on the same roads, same conditions. Complex intersections, tight curves, hills, merging traffic. All the fun stuff. Enough yapping.

 

[oktane]

The message is flawed.

So how, exactly, is granny depicted on your AP1 IC? Better yet, what, exactly does your AP1 car do with the pretty pictures it displays on your IC? If there’s no operational decision made based on a (sometimes incorrect, I’ve read) detected vehicle type, of what utility is it?

Maybe that’s why we have that nice, big screen with a drawing program; To appease those who like pretty pictures. Apparently it’s important to some.

Maybe identifying vehicles is the first step towards "making operation decisions based on detected vehicle type." Despite the logical fallacies in your attack towards another poster, I'm sure even you can understand that it is impossible to make "operational decisions based on detected vehicle type", when in fact, vehicle type is NOT identified.

Seeing an accurate representation of the traffic around the car depicted with correct vehicle types in correct adjacent lanes adds a certain level of "trust" to the AP, that in fact the car is aware of its surroundings. In the current AP2 iteration, seeing lanes bouncing around and lanes where non exist, and semi-trucks represented as small cars next to me is not confidence inspiring.

This is not a difficult concept to understand.

 

[Blup85]

Just a data point. I've put 350 miles on an ap2 mx100d loaner .50 firmware in the last 4 days while my .42 ap1 MS is getting work done.
I find the ap2 car constantly seeking and wandering while on the freeway a lot more than my ap1 ms. Depending on how many on/off exits there are in close proximity, it gets almost sickening in the far right lane while the car looks for the lane line. I also experienced my first random brake check ( 3 times), which really gives you an uncomfortable feeling and caused my wife to complain enough to turn off " death mode" (don't worry, she call my ap1 the same thing).
I also found acceleration to be a little more jerky when vehicles pull out of your lane allowing you to get to your higher set point.
I'm hoping the wandering is reduced when I finally find an MX that I can afford but In general Ap2 made the long journey a lot better than not having it, but I can't say it's better than ap1.

 

[buttershrimp]

It's ok to turn your phone sideways... won't hurt it a bit.

Ok, hold on here - since you’re local , do you mean divided highways to include only the 405, 110, 101, or to also include, say, Hawthorne or Sepulveda or some other boulevard with both medians and turn lanes as dividers (trying to come up with streets other than 2-lane meaning 1 lane in each direction - which neither of us meant).

I’ve been able to auto lane change fine heading up to SF via the 5 or 101, but not on the wider parts of PCH/Route 1. This is the same as what you’ve experienced, yes?

Also with regard to your previous question, AS especially with Chill on *is* much better with .48 compared to .42 relative to lane wandering. Problem is it still crosses lane markers and gets too close to adjacent traffic. Optimistically, once the other cameras are brought into play, the car should track pretty much perfectly.

I groan everytime I get sucked into the parity debate like it's a real thing... which is why I cringed when I read the tread spiral again after AR's post.... He is incredibly good at reframing the discussion in a way around a zero sum game... AP1 VERSUS AP2... One must DIIIIE.

I can respect the proof is the putting sort of thing in terms of a discussion about AP2 progress, but it's too simplistic and easy and lazy of an argument to pit AP1 against AP2. Then again, there is a reason why people flock to the binary movie plots over and over again with a different actor each time (Gerard Butler, Ben Affleck, Jeremy Renner, Wolverine, etc. etc.)...

The REAL intriguing debate is more along the lines of what @Blktesla's comments were regarding his discussion with autopilot insider and the central issue is creating a unified code set between versions for AP1,2 & 2.5!!!!

Now THAT is a much more accurate description of the software problem that must be solved! The consequences of this are that it muddies the waters and inadvertently pits AP1 against AP2. Sound like a dumb idea to do that? Yes! It creates a false comparison in so many ways. And yes, it sells movie tickets on the forum, but it's a waste of energy and brain power.

The more interesting question IMHO is to ask: "How much better would AP2 be right now if we didn't have to also make it work on AP1 cars? Imagine the possibilities if they could just walk away from making AP1 work? Should we? I don't know the answer, but damn it's interesting... and you sure as hell bet the people on the autopilot team think long and hard about this dilemma. Perhaps that is the real reason Lattner and Sterling didn't work out just as AP2 was taking off? Was it a Sophie's Choice situation?

The problem is that if you go down the rabbit hole too far.... you end up taking the Volvo approach. "Why try?" " Both AP versions suck." "Shame on Tesla for misleading us" sort of thing. Tesla haters on the forum all climax in unison. The Devil grins, and tell us, "money is on the dresser honey, go ahead and let yourself out!" Well, no thanks!!!!! You all can go stay at the Anxiety Ranger Hotel, but not me!

I'm not tired yet, and I have Autopilot rap videos to film for fun.

 

[⚡️ELECTROMAN⚡️]

One of my goals tomorrow is to compare anxiety ranger and TaoJones posts. Thinking maybe they are the same person. Probably aren’t, but I like frittering away my time in starange ways. Got a 7am Costco tire appointment with nothing else to do while I wait.