FSD Timeline Promises (summary)

[S4WRXTTCS]

What interesting there is if you look only at results from radar cars it's pretty close to even....while for vision only cars it's OVERWHELMINGLY bad.

I've mentioned before some concern about their getting vision-only to parity with radar+vision going slowly and this seems to reinforce that concern.


FWIW as someone with a radar car, who is now theoretically on vision only as part of FSDBeta, I still haven't noticed any real problematic "phantom" braking at all on NoA/Highway driving.


There's certainly lots of braking in city streets.... though it's way better in recent versions than when I first got it... Now it's more like reasonably often 1-3 mph slowdowns, pretty reliably if you crest a hill or come around a corner or anything else where visibility isn't ideal ahead.... whereas before it was a much harder slowdown.

There is definitely something wrong with pure vision.

Where I noticed the most degradation is in the rain. Even TACC will say "reducing speed due to weather", and Vision+Radar never did that. In fact the degradation is so pronounced that I really haven't had the opportunity to test out NoA with the FSD Beta build.

Can't really test NoA when it tells me the camera is too blinded to do an auto-lane change.

Hopefully I get a chance to test it more this week as the weather gods are giving us PNW folks a tiny break from the rain.

 

[S4WRXTTCS]

I wonder what will happen when they go to single stack. Will NoA continue to be basically a L3 system or will it be obviously L2 like FSD Beta?
I always figured FSD Beta was superset of NoA.

NoA under Enhanced Autopilot was initially intended as an L2 system, but after the rearrangement in 2019 it became became part of FSD.

So whatever ambition there is for FSD is the ambition for NoA. Now this doesn't mean they won't use flags in the SW to differentiate functionality differences between EAP owners, and FSD owners.

The biggest issue is probably going to be HW upgrades for HW2/HW2.5 EAP only owners, but I think Tesla will upgrade those people versus running some old stack. I can't see there being that many left in North America once FSD Beta really starts to provide functionality.

As to the ambition for FSD I'd say its at least L4.

L3 is simply something people want, but I haven't seen any indication from Elon/Tesla that they have any desire to enable that.

 

[daniel]

NoA under Enhanced Autopilot was initially intended as an L2 system, but after the rearrangement in 2019 it became became part of FSD.

So whatever ambition there is for FSD is the ambition for NoA. Now this doesn't mean they won't use flags in the SW to differentiate functionality differences between EAP owners, and FSD owners.

The biggest issue is probably going to be HW upgrades for HW2/HW2.5 EAP only owners, but I think Tesla will upgrade those people versus running some old stack. I can't see there being that many left in North America once FSD Beta really starts to provide functionality.

As to the ambition for FSD I'd say its at least L4.

L3 is simply something people want, but I haven't seen any indication from Elon/Tesla that they have any desire to enable that.

There's no way Tesla is going to upgrade the hardware in HW2/2.5 vehicles that have EAP but have not paid for FSD. With the proviso that it's L2 and that some of the (unimportant-to-me) features don't work, EAP does what was promised (mostly). In particular, it does highway lane-keeping and speed control excellently. There's no reason they'd have to upgrade EAP NoA to match FSD NoA.

If they succeed in producing real FSD there will have to be two separate software suites: One for FSD and one for AP/EAP. If you keep your phone or your computer long enough, there comes a time when you can no longer install the latest OS. The software for the FSD-track cars will already (I presume) not run on my car. That's to be expected.

 

[t3sl4drvr]

I wonder what will happen when they go to single stack. Will NoA continue to be basically a L3 system or will it be obviously L2 like FSD Beta?

NoA is L2: Driver is responsible for driving and expected to take over at any time driver sees need for it.

In L3, car is responsible for driving. Driver can keep their eyes closed and only need to take over if car requests so.

While they seem to be "close", they really are not. In one driver is responsible - it other the car is responsible. Lets not spread misinformation on NoA being L3 - in worst case, someone might start treating it as L3 and hurt themselves.

 

[95gator]

And that's why I say (and have been saying for years) that it is dishonest of Tesla to sell a feature (FSD) that does not exist. The fact is that until it exists we cannot know for sure if it ever will exist. I think it will eventually, but we cannot know. Tesla made promises it could not keep, and it is abundantly clear that the promise of being able to use your car as a robotaxi will not be fulfilled during the lifetime of any of the cars sold while Tesla was making that promise.

Selling something that does not exist is fraud. Selling a promise and then failing to make good on that promise is breach of contract. Some people bought FSD knowing it was a crap-shoot and were okay with that. But anybody who bought FDS because they believed Musk has been the victim of fraud.

Well I finally got 10.4 today with a 99 in cali which was NOT easy.

Sorry to all who have not. Not sure how I got it but not looking a gift horse in the mouth. Thank you Tesla!

It’s a bit daunting at first TBH but it’s so hella cool. I had my daughter (9 years) with me on the first drive and it occurred to me she will probably never drive a car the way we did.

Right?

Tesla is the Ford of Electric Cars. Ironically Ford is the Yugo of EVs.

 

[S4WRXTTCS]

There's no way Tesla is going to upgrade the hardware in HW2/2.5 vehicles that have EAP but have not paid for FSD. With the proviso that it's L2 and that some of the (unimportant-to-me) features don't work, EAP does what was promised (mostly). In particular, it does highway lane-keeping and speed control excellently. There's no reason they'd have to upgrade EAP NoA to match FSD NoA.

I'm not sure this is a no way.

The reason is EAP is a lot more than highway lane-keeping, and speed control.

It's Smart Summon
It's Auto Park
It's Auto Lane Change
It's emergency vehicle detection on the highway that they just added, but did they add it for HW2/HW2.5 owners?
It's a future pot hole detection on the highway
It's a future offset vehicle detection/avoidance on the highway

Now sure Tesla has an existing stack that they can claim matches what was promised, but there is a cost to not upgrading them.

It's far more likely that a customer will subscribe to FSD if they don't have pay the $1K fee to upgrade
It's far more likely that a satisfied customer will subscribe to FSD than a customer that doesn't even think Tesla can get EAP working let along FSD.
If there are safety related aspects improvements than it helps get the NHTSA off your back.

I do expect Tesla to continue down the "you pay for it" path, but at some point they might upgrade them.

 

[S4WRXTTCS]

Well I finally got 10.4 today with a 99 in cali which was NOT easy.

Sorry to all who have not. Not sure how I got it but not looking a gift horse in the mouth. Thank you Tesla!

It’s a bit daunting at first TBH but it’s so hella cool. I had my daughter (9 years) with me on the first drive and it occurred to me she will probably never drive a car the way we did.

Right?

Tesla is the Ford of Electric Cars. Ironically Ford is the Yugo of EVs.

I'm disappointed that you didn't do a comparison between your 9 year old driving, and 10.4 driving.

 

[S4WRXTTCS]

If they succeed in producing real FSD there will have to be two separate software suites: One for FSD and one for AP/EAP. If you keep your phone or your computer long enough, there comes a time when you can no longer install the latest OS. The software for the FSD-track cars will already (I presume) not run on my car. That's to be expected.

I think it depends on the HW level where they succeed.

If they succeed with HW3 I think they'll have legacy SW for older HW, and a single software suite for FSD/EAP/AP running on HW3.

Now I don't actually think they'll succeed with HW3, and it will require significantly different HW. Where there won't be an easy upgrade path so Tesla will find some other way to try to make FSD owners whole. Hopefully it will be lifetime FSD subscription or something like that. I completely agree with you in that you can only expect SW upgrades to work for so long on your current device. Tesla already threw out hundreds of thousands of HW2.5 computers. The last thing we want is them throwing out HW3 computers, HW4 computers, etc.

 

[95gator]

I'm disappointed that you didn't do a comparison between your 9 year old driving, and 10.4 driving.

Want some peanut butter with your jelly?

Seriously though, it drives way better on the streets than the highway, which is where I usually drive. Didn’t know it was gonna effect AP and I have a long drive coming up.

 

[daniel]

I'm not sure this is a no way.

The reason is EAP is a lot more than highway lane-keeping, and speed control.

It's Smart Summon
It's Auto Park
It's Auto Lane Change
It's emergency vehicle detection on the highway that they just added, but did they add it for HW2/HW2.5 owners?
It's a future pot hole detection on the highway
It's a future offset vehicle detection/avoidance on the highway

Now sure Tesla has an existing stack that they can claim matches what was promised, but there is a cost to not upgrading them.

It's far more likely that a customer will subscribe to FSD if they don't have pay the $1K fee to upgrade
It's far more likely that a satisfied customer will subscribe to FSD than a customer that doesn't even think Tesla can get EAP working let along FSD.
If there are safety related aspects improvements than it helps get the NHTSA off your back.

I do expect Tesla to continue down the "you pay for it" path, but at some point they might upgrade them.

In spite of all the talk here about how much alike EAP and FSD are supposed to be, they are entirely different concepts. EAP is and always was intended to be Level 2. Some of what it is supposed to do is analogous to what FSD aspires to be, but the fundamental difference is that with EAP the car bears no responsibility and with FSD (if it ever exists) the driver bears no responsibility. And that makes all the difference.

EAP does not have to work perfectly or all the time in order to meet its promises. FSD has to work nearly perfectly (better than human) all the time under every conceivable circumstance. It does not require a hardware upgrade for EAP to meet its promise because the promise was never perfection.

I certainly would not refuse an upgrade to HW3 and if they offered me FSD for free I'd accept it (though I would not use it until I read better reviews than I'm reading now). But OTOH, if they succeed in producing true non-geofenced Level 4 FSD I will trade in my car for it and pay whatever they decide to charge.

Do they still need to take cars into a service center to replace the computer? Because if they do, that would mean shipping my car to Oahu. There's no SC here. They'd have to pay a pretty penny to ship my car AND provide me with a loaner and it would likely take at least two weeks for the round-trip shipping.

I paid for EAP and I'm very happy with it. No new computer needed. I don't believe they're going to upgrade my hardware just with some notion that EAP should function at a higher level than it does.

 

[S4WRXTTCS]

In spite of all the talk here about how much alike EAP and FSD are supposed to be, they are entirely different concepts. EAP is and always was intended to be Level 2. Some of what it is supposed to do is analogous to what FSD aspires to be, but the fundamental difference is that with EAP the car bears no responsibility and with FSD (if it ever exists) the driver bears no responsibility. And that makes all the difference.

EAP does not have to work perfectly or all the time in order to meet its promises. FSD has to work nearly perfectly (better than human) all the time under every conceivable circumstance. It does not require a hardware upgrade for EAP to meet its promise because the promise was never perfection.

I certainly would not refuse an upgrade to HW3 and if they offered me FSD for free I'd accept it (though I would not use it until I read better reviews than I'm reading now). But OTOH, if they succeed in producing true non-geofenced Level 4 FSD I will trade in my car for it and pay whatever they decide to charge.

Do they still need to take cars into a service center to replace the computer? Because if they do, that would mean shipping my car to Oahu. There's no SC here. They'd have to pay a pretty penny to ship my car AND provide me with a loaner and it would likely take at least two weeks for the round-trip shipping.

I paid for EAP and I'm very happy with it. No new computer needed. I don't believe they're going to upgrade my hardware just with some notion that EAP should function at a higher level than it does.

Does EAP on HW2.X work well enough that the majority of buyers are happy with it? Where the features, and the performance of those features match expectations?

EAP is plagued by four things.

It's plagued by features like Smart Summon, and Autopark being part of it. But, neither feature really has anything to do with L2 highway driving. For these feature to work they require a lot of processing power.

It's plagued by Tesla ignoring HW2/HW2.5 while all the development work happens on HW3.

It's plagued by HW2.X being such a small part of the overall fleet. Why optimize for an architect that so few people have?

It's plagued by Tesla using so much of HW3 for FSD that they even if they manage to get it to work it might only work at L2 because of a lack of redundancy. The EAP at L2, and FSD at an autonomous level was supposed to make all the difference. It was supposed to allow EAP to work on lesser HW. But, now the lesser HW is looking to be HW3.

For full disclosure the primary reason I paid $3K for FSD (on top of the $5K for EAP) is the feeling like old HW would be forgotten about, and that ultimately I was better off with HW3 that was promised with FSD. My expectation was that they wouldn't be able to pull off Autonomous driving, but that NoA would be pretty sweet. It's been over 3 years and NoA still doesn't work that well.

Like Auto-lane change has roughly a 50% chance of canceling if its attempted when a semi-trailer is the lane next to the one your changing into. At least that's been my experience with it.

 

[daniel]

Does EAP on HW2.X work well enough that the majority of buyers are happy with it? Where the features, and the performance of those features match expectations?

EAP is plagued by four things.

It's plagued by features like Smart Summon, and Autopark being part of it. But, neither feature really has anything to do with L2 highway driving. For these feature to work they require a lot of processing power.

It's plagued by Tesla ignoring HW2/HW2.5 while all the development work happens on HW3.

It's plagued by HW2.X being such a small part of the overall fleet. Why optimize for an architect that so few people have?

It's plagued by Tesla using so much of HW3 for FSD that they even if they manage to get it to work it might only work at L2 because of a lack of redundancy. The EAP at L2, and FSD at an autonomous level was supposed to make all the difference. It was supposed to allow EAP to work on lesser HW. But, now the lesser HW is looking to be HW3.

For full disclosure the primary reason I paid $3K for FSD (on top of the $5K for EAP) is the feeling like old HW would be forgotten about, and that ultimately I was better off with HW3 that was promised with FSD. My expectation was that they wouldn't be able to pull off Autonomous driving, but that NoA would be pretty sweet. It's been over 3 years and NoA still doesn't work that well.

Like Auto-lane change has roughly a 50% chance of canceling if its attempted when a semi-trailer is the lane next to the one your changing into. At least that's been my experience with it.

I cannot speak for "the majority of buyers," or for anyone but myself, for that matter. But I am delighted with EAP. I have one major complaint, which I'll get to below.

I bought it for autosteer and TACC. I'd have bought plain AP but when I got my car the choices were: 1. Conventional cruise control, no autosteer; 2. EAP; 3. FSD, which at the time was EAP and the Big Empty Promise of fully driverless in the future. There are no freeways here; autopark does not work for me and I don't care; I've never tried summon. Autosteer works wonderfully, and TACC has actually improved since I got the car and is now wonderful. (It had an issue with excessive lag at first.) Lane change from the turn-signal stalk works some of the time. Not a big deal for me. I seldom change lanes when using autosteer.

It does not bother me that Tesla has moved on to its long-term goal of FSD and the new hardware. My car does what I paid for it to do.

My complaint: Autosteer will not operate faster than 5 mph over the posted limit. On Maui speed limits are idiotically slow and everybody drives 10 to 15 mph over the limit. So when there is a lot of traffic I cannot use autosteer without being a public nuisance and holding back a very long line of cars, and creating an actual safety hazard as they all try to pass me. So when the traffic is heavy I do not use autosteer. I wish this limitation did not exist.

 

[Knightshade]

My complaint: Autosteer will not operate faster than 5 mph over the posted limit. On Maui speed limits are idiotically slow and everybody drives 10 to 15 mph over the limit. So when there is a lot of traffic I cannot use autosteer without being a public nuisance and holding back a very long line of cars, and creating an actual safety hazard as they all try to pass me. So when the traffic is heavy I do not use autosteer. I wish this limitation did not exist.

It doesn't exist if you use it where it's intended to be used (limited access divided highways per the manual)

You're using it outside the ODD, so it has further limitations to the system (also per the manual).

City streets (in FSDBeta anyway), FWIW, has no such limitations and you can set the speed to anything you want up to the max (80) regardless of posted speed limit- same as you can regular EAP on freeways.

Does EAP on HW2.X work well enough that the majority of buyers are happy with it? Where the features, and the performance of those features match expectations?

Yup.

Longer range on smart summon would be nice I guess... but overall quite happy with it.

EAP is plagued by four things.

Pretty sure you're not using "plagued" correctly here.

It's plagued by features like Smart Summon, and Autopark being part of it. But, neither feature really has anything to do with L2 highway driving.

Yeah... "does more stuff than you'd expect" isn't really a plague.

Heck even if the "extra" stuff isn't great that's not a plague.

For these feature to work they require a lot of processing power.

<citation needed>

Pretty sure autopark uses a lot less processing power than Navigate on Autopilot for example.... (and it's not like both need to run at the same time)

It's plagued by Tesla ignoring HW2/HW2.5 while all the development work happens on HW3.

...what?

EAP has been feature complete since 2019. How are they "ignoring" something that's already done?

It's plagued by HW2.X being such a small part of the overall fleet. Why optimize for an architect that so few people have?

Again- not a "plague"

HW2s promised features have all been delivered. 2+ years ago. It's done. They'll fix bugs and occasionally clean up code, but it's not getting anything "new" anymore. Same as HW1 was years ago.

If you like what you've got- great.

If you want other, new, stuff- you'll need to buy it (that's FSD- and it comes with free hardware upgrades as needed too!)

It's plagued by Tesla using so much of HW3 for FSD that they even if they manage to get it to work it might only work at L2 because of a lack of redundancy.

...wait, what?

EAP is "plagued" by how much compute stuff that is not part of EAP uses?

EAP, a system never intended to be more than L2 is "plagued" by being... L2 exactly as ever promised?

Yeah- you're definitely using that word wrong.

The EAP at L2, and FSD at an autonomous level was supposed to make all the difference. It was supposed to allow EAP to work on lesser HW. But, now the lesser HW is looking to be HW3.

I'm not even sure what that sentence is trying to say.

EAP is L2. It only needs HW2.

And it's been feature complete since ~October 2019.

It doesn't "need" anything. It's finished.

For full disclosure the primary reason I paid $3K for FSD (on top of the $5K for EAP) is the feeling like old HW would be forgotten about, and that ultimately I was better off with HW3 that was promised with FSD.

Same.

My own expectation was with a computer upgrade I could probably get L3 highway, and I'd be perfectly happy with that (since at the time 95% of my driving was highway and it'd be nice to read a book or something while doing that)

I still think the sensor suite is sufficient for that task. Single stack when it comes to highways will give me a much better idea how near the software is for it though--- as from what I've seen of FSDBeta it already has code to address most if not all of what's "missing" for it to be L3 highway right now (dealing with partly-in-lane objects and handling temporarily vanishing lane lines)

My expectation was that they wouldn't be able to pull off Autonomous driving, but that NoA would be pretty sweet. It's been over 3 years and NoA still doesn't work that well.

Works awesome for me. Still use it for nearly all my driving, virtually never have to intervene for anything.

Any time I'm forced to driving anything else like my wifes vehicle I'm like "What? I have to STEER, like some sort of peasant??"

Like Auto-lane change has roughly a 50% chance of canceling if its attempted when a semi-trailer is the lane next to the one your changing into. At least that's been my experience with it.

Not remotely mine.

It could use some improvement to the logic on WHEN to change lanes based on upcoming exists, but I can't recall the last time it aborted an auto lane change (it did this fairly often when the feature first came out but not for a long time)- and it hasn't aborted an automatic-but-user-prompted one (which it does significantly quicker) in years.

 

[S4WRXTTCS]

Pretty sure you're not using "plagued" correctly here.

It comes down to expectations.

If you're tormented because you can't achieve the expectations placed on you than you're plagued by those expectations.

I listed the reasons why EAP would really never achieve what was promised. I agree that its technically feature completely, but I think it leaves a lot to be desired in terms of how well it performs.

EAP could have been something special, but the burden placed on it was just too high. Then it suffered the worst fate of being axed. It's one last breath was when Tesla brought it back for a month for a discount.

As to auto-lane change I welcome any input from PNW'er with their experience on I5 with semitrucks. I rarely have a cancelation unless a semi-truck is in the lane next to the one I'm changing into. Usually this is getting out of the passing lane, and into the middle lane. I don't use the unconfirmed lane changes so these are all ones I initiated. This happens on a 200+ mile to Portland so there are quite a few lane changes as there are a ridiculous amount of semi-trucks on that freeway with a much slower speed limit than cars.

Out of all the features of EAP the auto-lane change is certainly my favorite.

 

[Knightshade]

It comes down to expectations.

No, it comes down to the actual meaning of the word, which you're misusing.

If you're tormented because you can't achieve the expectations placed on you than you're plagued by those expectations.

This reminds me of the super drama in the HW2.0 FSD thread where people were screaming about their SUFFERING because they had to wait a month for free camera upgrades.

It's nonsensical.

EAP told you pre-purchase what it would do. It does it.

I listed the reasons why EAP would really never achieve what was promised.

And I debunked them.

I agree that its technically feature completely, but I think it leaves a lot to be desired in terms of how well it performs.

Maybe there's something wrong with your car?

I just drove, my house, to my workplace, over 35 miles, and the system did all the work. (Including the couple miles non-highway on both ends, but that's city streets, so EAP only gets the credit for about 90% of that trip).

EAP could have been something special

It is.

It's the reason I bought the car, and the reason I hate driving any non-tesla for lack of it.

, but the burden placed on it was just too high.

This again is your imagination at work.

If YOU expected a bunch of magical stuff EAP never promised, that's a failure on your part, not EAPs.

 

[DarkandStormy]

https://twitter.com/x/status/1333855145772376066

One year and one day since Elon said this.

 

[EVNow]

One year and one day since Elon said this.

4 Years since GM/Cruise said this.

GM plans large-scale launch of self-driving cars in U.S. cities in 2019

General Motors Co laid out its vision for self-driving vehicles on Thursday, telling investors it planned a commercial launch of fleets of fully autonomous robo-taxis in multiple dense urban environments in 2019, in a challenge to rivals such as Alphabet Inc's Waymo.

www.reuters.com

General Motors Co laid out its vision for self-driving vehicles on Thursday, telling investors it planned a commercial launch of fleets of fully autonomous robo-taxis in multiple dense urban environments in 2019, in a challenge to rivals such as Alphabet Inc’s Waymo.

 

[Daniel in SD]

4 Years since GM/Cruise said this.

They should give refunds to the customers they sold these “robotaxis” to.

 

[EVNow]

They should give refunds to the customers investors they sold these “robotaxis” to.

FTFY.

 

[Daniel in SD]

FTFY.

Forward-looking statement - Wikipedia

en.m.wikipedia.org

Elon’s talking about a future product called FSD which is perfectly fine except that Tesla sells a different product that’s also called FSD.