Will AP 2 ever be “Upgraded” to AP 1.0?

[xyeahtony]

AP2.0/2.5 does a lot more than AP1.0.

The only thing AP1.0 does that AP2.0 doesn't is that it shows semis and bikes and people avatars on the AP screen, while AP2.0 just shows everyone else as a tesla sedan.

Why do people doubt AP2>AP1? It's simple logic. 8 cameras > 2 cameras. Technically one camera on AP1 since they dont use the rear view camera.

 

[ℬête Noire]

No as AP 2/2.5 will be upgraded to AP 3.0 if you purchased FSD. Jim Keller designed the chip. This is the same guy that designed Apple iPhone CPUs, AMD Athlon and AMD Zen/Ryzen, cofounder of the x86-64 instruction set, Hypertransport and even the old DEC Alpha CPUs. He tends to change jobs after completing a design. He has worked for Apple, AMD, Tesla and now Intel.

Everything he touches is gold so I purchased FSD on my order after hearing about AP 3.0.

It did indeed give me pause for thought about putting down the $3000, Tesla might actually be able to have enough imaging processing power to use the complete camera complement as an optic spectrum stand-in for LIDAR.

However weren't they supposed to roll out something of a feature this month for FSD? I know the new processor is in the future but I'm still not ready to put down non-refundable money for vapour.

 

[xyeahtony]

It did indeed give me pause for thought about putting down the $3000, Tesla might actually be able to have enough imaging processing power to use the complete camera complement as an optic spectrum stand-in for LIDAR.

However weren't they supposed to roll out something of a feature this month for FSD? I know the new processor is in the future but I'm still not ready to put down non-refundable money for vapour.

Software version 9. More EAP features. If you plan on keeping your car for 5+ years, FSD will eventually come. Its not an IF, but when.

LIDAR arguments aside, i've never really understood that as humans don't use "lidar" and we can perceive everything optically. Cameras can do that even better. The sensor suite is all there for FSD, we just need the software.

 

[ℬête Noire]

Software version 9. More EAP features. If you plan on keeping your car for 5+ years, FSD will eventually come. Its not an IF, but when.

Someone tells you "put $3000 in a hole in the ground and you'll get it back in maybe 5 years...as long as the area doesn't have a flood that washes the whole thing away" The later in this analogy that's a car crash or something else unforeseen leaving you without the vehicle. That seems a very sketchy financial decision on the face of it. You can have reasons for doing it ("I want to fund the effort to get AI driving faster") but in the scope of being "worth it" the risk-reward vs paying the extra $2000 or whatever down the road isn't that good.

LIDAR arguments aside, i've never really understood that as humans don't use "lidar" and we can perceive everything optically. Cameras can do that even better. The sensor suite is all there for FSD, we just need the software.

"Just need the software" reads like "just needs magic" here. Humans have highly specialized processing gear...and they make errors. AI driving needs to be better than nearly all humans or it's pretty much useless.

The strength of active sensing (what LIDAR is) is that its data output is a better fit for the computer. It doesn't need to worry about the multiple bandwidth (it is monochrome), and it's processing one "point" at a time, meaning a couple less degrees of freedom for noise to happen on. That makes it easier to be certain about the model of the real world it's building.

They are going to need to suss out the 3D data from binocular camera views to replace that.** Which they just can't do right now.

** Humans don't really rely on this much, you can have a unrestricted driver's license with only one eye. You can fairly quickly adopt to a lot of tasks with only one eye. The second eye really only fills in holes from the first, it's not that critical in judging depth as many people think.

 

[Knightshade]

AP2.0/2.5 does a lot more than AP1.0.

The only thing AP1.0 does that AP2.0 doesn't is that it shows semis and bikes and people avatars on the AP screen, while AP2.0 just shows everyone else as a tesla sedan.

Why do people doubt AP2>AP1? It's simple logic. 8 cameras > 2 cameras. Technically one camera on AP1 since they dont use the rear view camera.

Quick nitpick- AP2+ doesn't use 8 cameras.

It uses a max of 4, and was only using 2 until recently IIRC

FSD will use all 8

 

[Knightshade]

Someone tells you "put $3000 in a hole in the ground and you'll get it back in maybe 5 years...as long as the area doesn't have a flood that washes the whole thing away" The later in this analogy that's a car crash or something else unforeseen leaving you without the vehicle. That seems a poor financial decision..

Some of the features that 3k gets you are coming this month (to early testers) and next month to everyone else.

We're unclear which FSD features they are though.

that's not "maybe 5 years"

It's not full FSD either, but it'll be the first time you've actually getting some tangible functionality for the $3000.

 

[eye.surgeon]

I've been using AP1 since day 1. It was actually much better then. It was at it's peak about 6 months after launch--worked on all roads smoothly and no nags. Now it only works on divided highways and since the database is crap that means it actually won't work on many divided highways. It's speed restricted to 5 over the speed limit, and since the database is crap that means is actually is speed limited to a variety of erroneous incorrect speed limits. The nags are ridiculous, pervasive, and render it much less useful and capable than it used to be.

 

[xyeahtony]

\

The strength of active sensing (what LIDAR is) is that its data output is a better fit for the computer. It doesn't need to worry about the multiple bandwidth (it is monochrome), and it's processing one "point" at a time, meaning a couple less degrees of freedom for noise to happen on. That makes it easier to be certain about the model of the real world it's building.

Lidar is not the ultimate solution people make it out to be. it is capable of mistakes as well. See Uber/Arizona crash. Lidar is mainly limited by poor visibility conditions that a camera suite can overcome. (fog anyone?)

tesla clearly believes the camera suite is adequate. they've staked their future and reputation on it. if there was a hair of doubt, i dont think they would've committed to putting the cameras into every single car they make if they didn't think it would have a purpose.

 

[ℬête Noire]

Lidar is not the ultimate solution people make it out to be. it is capable of mistakes as well. See Uber/Arizona crash. Lidar is mainly limited by poor visibility conditions that a camera suite can overcome. (fog anyone?)

The LIDAR sub-system was successful there. The pedestrian was detected in time to likely have averted the death if not the collision outright. That death was due to a moral failure in what the people at Uber decided to do about the input (ignore it because it could cause vehicle passengers to be discomforted by hard braking).

tesla clearly believes the camera suite is adequate.

They aren't using the suite. Please re-read my initial post you replied to.

 

[Saghost]

I've been using AP1 since day 1. It was actually much better then. It was at it's peak about 6 months after launch--worked on all roads smoothly and no nags. Now it only works on divided highways and since the database is crap that means it actually won't work on many divided highways. It's speed restricted to 5 over the speed limit, and since the database is crap that means is actually is speed limited to a variety of erroneous incorrect speed limits. The nags are ridiculous, pervasive, and render it much less useful and capable than it used to be.

Disagree. Maybe not with the nags are annoying part, but with the better at sux months part definitely.

I came in to this game at 7.1 two years ago, and my AP1 X is a much better driver than it used to be, with things like truck lust and always trying to exit largely eliminated and AP handling bad weather and shadows on the road and poor lane lines much more effectively.

Not good enough to trust without paying periodic attention, but much better than it was.

 

[RDoc]

I've been using AP1 since day 1. It was actually much better then. It was at it's peak about 6 months after launch--worked on all roads smoothly and no nags. Now it only works on divided highways and since the database is crap that means it actually won't work on many divided highways. It's speed restricted to 5 over the speed limit, and since the database is crap that means is actually is speed limited to a variety of erroneous incorrect speed limits. The nags are ridiculous, pervasive, and render it much less useful and capable than it used to be.

It works fine for me on local roads, although with the speed restriction. It's particularly useful in stop and go situations.
My AP1 reads speed signs, almost always correctly, and I've never had it restrict my speed on a highway.
By simply holding the wheel, I never get nags.

 

[sdrevik]

Thanks again to everyone for the correction. I had seen some thread where someone had asked why is the AP icon Would not show that auto pilot was available, and two people responded “only on limited access highway is with a clear would not show that auto pilot was available, and two people responded “only on limited access highway’s with a clear median”. It appears they should have added “if you are going over the speed limit.”

Glad to hear!

 

[Atari2600]

Someone tells you "put $3000 in a hole in the ground and you'll get it back in maybe 5 years...as long as the area doesn't have a flood that washes the whole thing away" The later in this analogy that's a car crash or something else unforeseen leaving you without the vehicle. That seems a very sketchy financial decision on the face of it. You can have reasons for doing it ("I want to fund the effort to get AI driving faster") but in the scope of being "worth it" the risk-reward vs paying the extra $2000 or whatever down the road isn't that good.
.
.
.
They are going to need to suss out the 3D data from binocular camera views to replace that.** Which they just can't do right now.

I want to fund the effort to get AI driving faster. Plus I want the new processor in 6 months to a year. I plan on keeping the car for 6 years. If the car is destroyed fine that $3K was in my original purchase price of the car and NOT $5k of add-ons after that fact.

You are correct and I assume the 3D data will only be forward viewing in conjunction with the radar and ultrasonic sensors. I am concerned there may not be enough data there for blind spot detection.

 

[m3tx]

Some of the features that 3k gets you are coming this month (to early testers) and next month to everyone else.

We're unclear which FSD features they are though.

that's not "maybe 5 years"

It's not full FSD either, but it'll be the first time you've actually getting some tangible functionality for the $3000.

The last tweet from Elon on that topic was that the first FSD features would come at same time as the chip upgrade, i.e. not until 6 months from now at the earliest.

 

[Knightshade]

The last tweet from Elon on that topic was that the first FSD features would come at same time as the chip upgrade, i.e. not until 6 months from now at the earliest.

So he contradicted this one where he says August? (which they later said was for test vehicles and September for everyone else)

Elon Musk on Twitter

 

[MXWing]

I apologize, I don’t have my M3 yet (Monday), but I had read on *numerous* current threads that current AP only works on highways with divided medians (and nobody is correcting those threads).

What am I missing with the contradiction in the threads on TMC?

My bad, the OP was based on what other people were telling me on other threads. Apologies, again.

What thread and what posters.

A good amount of users on TMC probably don't even own a Tesla let alone used Autopilot.

You get straight talk here based on the responses I seen so far.

Much love to taking you well deserved beating like a man.

Enjoy your Model 3 and AP2.5

 

[MIT_S60]

I spent a fair amount of time with two AP1 service loaners this month, and I own an AP2.0 car (since before any features were activated). Here's my breakdown:

AP1:
-Seemed to engage in following a car ahead without seeing road lanes much better than AP2.0
-Reads speed limit signs (AP2.0's speed limit data is constantly off)
-Sees motorcycles/trucks
-Allows lane change on local roads (AP2.0 only allows lane change on freeways)

AP2.0:
-Slows down for curves on ramps between freeways (AP1.0 seems to try to go the full speed, which led me to disengage AP in those scenarios).
-Appeared to "see" lanes better on local roads.

I also found both AP1 service loaners did a MUCH better job about basically never nagging me, whereas my AP2.0 car nags me ALL the time (with one or two hands on the steering wheel).

 

[burnside]

Just my observation, but using AP on local roads will sometimes have issues with larger intersections since they do not have lane markers. Might be a reason why AP is discouraged on local roads.

 

[m3tx]

So he contradicted this one where he says August? (which they later said was for test vehicles and September for everyone else)

Elon Musk on Twitter

Yep, he sure did
Elon Musk on Twitter

 

[xav-]

AP2.0/2.5 does a lot more than AP1.0.

The only thing AP1.0 does that AP2.0 doesn't is that it shows semis and bikes and people avatars on the AP screen, while AP2.0 just shows everyone else as a tesla sedan.

Why do people doubt AP2>AP1? It's simple logic. 8 cameras > 2 cameras. Technically one camera on AP1 since they dont use the rear view camera.

“people avatar not displayed”

You should really be saying instead “there is no pedestrian detection” on AP20 car. This is not just a display issue.. and that is a big thing, regardless of how many cameras you use.