-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Enhanced Summon, where are you?
- Thread starter dckiwi
- Start date
anthonyj
Stonks
Just rolled through that stop sign not giving two shits. Just like an average person would. I'd say it's ready
EVNow
Well-Known Member
This is a good example of trolling.
AlanSubie4Life
Efficiency Obsessed Member
Yeah, that is kind of a crazy limitation - in addition it appears to align with the way those cars are parked. But see my picture above! It worked with one car!
Anyway, back to Enhanced Summon...just wait, make it good, remove the "Really" verbiage from the website if necessary, and keep on developing the software. There is no rush. The highest quality entry will ultimately win (just as is proving to be the case on the vehicle hardware side, when it comes to the Model 3). Settling lawsuits and such from unsatisfied customers is just noise. Now, make the software good, do what is technically feasible with HW2.5, and let the chips fall where they may. (Personally I don't think there should be a huge difference between HW2.5 and HW3 at parking lot speeds, though the amount of image resolution available might matter I suppose. Obviously there are huge hardware capability differences, but I've heard zero mention about limitations due to that hardware in regards to Smart/Enhanced Summon)
This is a hard problem for reasons quite distinct from the hardware! If they can't address these issues, FSD is a LONG way off. Enhanced Summon is a minor feature and not worth obsessing over and constantly tweeting about the timeline.
Release version 10 without Enhanced Summon when it is ready - plenty of features & enhancements there to help drive additional sales without a minor feature such as Enhanced Summon.
Actually within the context of this topic ”Enhanced Summon, where are you?” my only point is this:
I think it makes a certain amount of sense for Tesla — and indeed it looks to me as if Tesla is in the process of doing just that — to wrap up AP2/2.5, as it has become clear they will not have sufficient power to go where Tesla wants to go software 2.0 wise.
So to wrap up AP2/2.5, Tesla needs to deliver on what was promised: EAP (Enhanced Autopilot). I think we can ignore the AP1 parity on traffic sign recognition, because I think Tesla may just forget about that one as it was not in the Design Studio but an indirect obligation based on a verbal comment... and they could say it has been superceded by database and Fleet Speed etc.
So what is missing? Enhanced Summon as it was on the AP2 website. They need to deliver that missing piece (and it helps that they can probably recognize some revenue out of it).
Once NoA and Enhanced Summon are delivered to most markets (where legal), EAP is done and Tesla can start to mothball AP2/2.5 like they have done with AP1, small maintenance updates only, and focus on FSD on HW3 (retrofitted HW2/2.5 included) which will allow them to take the next step in how the autonomous dream is pursued. I think they will leave AP2/2.5 in perpetual beta, like AP1.
At this point HW3 will start running new NNs etc and the EAP fleet with the old, simpler logic etc may be left behind. EAP cars will have the old Summon, old Enhanced Summon, old Autopark. Eventually FSD on HW3 will solve the parking lot, all this and more in new and better ways, is the hope.
Now... when does one believe FSD on HW3 will achieve all the promised things on Autonomy Investor Day? Another thread maybe.
Last edited:
jebinc
Well-Known Member
Interesting perspective. Being a relatively new Tesla owner, I'm not 100% up to speed on promises of yesteryear, or what's in the bowls of AP1, AP2/2.5 software and/ or there limitations. Having said that, it sounds like you believe all software released today was/is optimized for the non HW3 platforms. Yes, I know it all "runs" on all non HW3 platforms, but does that mean cars running HW3 run this software less optimally (kind of like a bad port - Think of that retro tiny PS1 console that was released just before the holiday's last year for $99 - Boy was that a botch job)? I suppose, since the HW3 computer is a native Tesla design and not Nvidia, that probably is the case. The question is, has Tesla even started any software development optimized for HW3? If not, we are in for a loooooong dry spell...Actually within the context of this topic ”Enhanced Summon, where are you?” my only point is this:
I think it makes a certain amount of sense for Tesla — and indeed it looks to me as if Tesla is in the process of doing just that — to wrap up AP2/2.5, as it has become clear they will not have sufficient power to go where Tesla wants to go software 2.0 wise.
So to wrap up AP2/2.5, Tesla needs to deliver on what was promised: EAP (Enhanced Autopilot). I think we can ignore the AP1 parity on traffic sign recognition, because I think Tesla may just forget about that one as it was not in the Design Studio but an indirect obligation based on a verbal comment... and they could say it has been superceded by database and Fleet Speed etc.
So what is missing? Enhanced Summon as it was on the AP2 website. They need to deliver that missing piece (and it helps that they can probably recognize some revenue out of it).
Once NoA and Enhanced Summon are delivered to most markets (where legal), EAP is done and Tesla can start to mothball AP2/2.5 like they have done with AP1, small maintenance updates only, and focus on FSD on HW3 (retrofitted HW2/2.5 included) which will allow them to take the next step in how the autonomous dream is pursued. I think they will leave AP2/2.5 in perpetual beta, like AP1.
At this point HW3 will start running new NNs etc and the EAP fleet with the old, simpler logic etc may be left behind. EAP cars will have the old Summon, old Enhanced Summon, old Autopark. Eventually FSD on HW3 will solve the parking lot, all this and more in new and better ways, is the hope.
Now... when does one believe FSD on HW3 will achieve all the promised things on Autonomy Investor Day? Another thread maybe.
boonedocks
MS LR Blk/Blk 19”
What I would hope is that there will, shortly, be a time where non FSD functions will be pretty much settled and much more of Tesla's resources can be targeted at HW3 FSD since HW3 will be the only way to have FSD. I am sure there will be minor tweaks and fixes as there are to AP1 but surely most of the braintrust resources will new targeted at HW3 only.
Obviously I ”know” nothing about Tesla’s private plans so this is all just theory I am toying with.
So, here’s what I think would be the case if this theory proves correct:
I believe all current software, including the one that actually runs on HW3, is basically optimized for the lowest common denominator that is HW2 (with a few HW2.5+ features) and runs fairly similarly well. But in my theory Tesla will stop developing this current software fairly soon, once EAP is delivered. After this they would merely provide maintenance type of updates for EAP as well as HW2/2.5 in general.
And that is what HW2/2.5 would in my theory run for all eternity. Once Tesla has delivered on its software promises to HW2/HW2.5 that would be the end of that. Exception would be HW2/2.5 buyers that buy FSD and get HW3 computer retrofitted, after which they would fall into the HW3 category in this theory.
At this current stage HW3 too runs this ”soon obsolete” software — or a similar variant that is slightly adjusted for HW3 as there are of course already some lower level differences today. So everything in public runs ”the same stuff” at the moment (except for the token HW2.5+ only features of course).
The difference is: Unlike HW2/2.5, HW3 would not run this codebase for all eternity. Once the ”old world” is wrapped up, HW3 cars would start introducing new stuff that will eventually replace the old codebase, perhaps entirely... similar to how AP2 was initially a sort of bad clone of AP1 but eventually went in a different direction. HW3 at some point would start replacing HW2/2.5 code with completely new stuff that would not be backported to HW2/2.5.
It is just a theory.
jebinc
Well-Known Member
So, in conclusion, "no developed and optimized for HW3" code arriving in our M3's anytime in the foreseeable future; and certainly not "Feature complete" and running on HW3 by year's end (2020 even). That's the take-away for me, "IF" your "hypothesis" is correct.
Do you have to continuously hold the summon button down to do this? What do you do if you saw a car start to back out and wanted to stop it while it was on it's own?
I am not sure that necessarily follows. Let’s assume for a moment that still during September Tesla releases Enhanced Summon and calls ”Mission Accomplished” on EAP...
They would then be ”free” to focus on HW3 and FSD (and HW2/2.5 retrofits) as the HW2/2.5 would be ”done”. They may well have some NNs and code ready to roll for that in the wings that they have been holding back because of HW2/2.5 compatibility.
This has essentially been my line of thinking as well. I don’t think it’s too far off base. Also explains why smart summon is taking so long and will ultimately be an entirely different beast with FSD computer and future software and NN training.
jebinc
Well-Known Member
Yes, that was the question - Concurrent/dual development (with singular release) or serial development... I hope it's the former and not the later....
I’m sure there is some concurrent development but I agree we must tread carefully here. Back in 2016-2017 there was a wide belief Tesla had a separate ”FSD codebase” that would replace the crappy EAP codebase any day now... well, never happened of course and Tesla’s AP team and solution has been redone a couple of times since if we know correct...
Do you not think that they are working on updated HW3 to HW3.5 remember HW3 is built under a Dual architecture its first so its a beta only Tesla knows how far they are from HW3.5 then the question is when do they initiate software optimization to hardware needs.
jebinc
Well-Known Member
No, this is a good example of you not having a clue - based on most recent posts, it is you it seems that haven't heeded the advice of other TMC members, that is to knock it off beating on other people while you push your agenda, but rather focus on the specific thread issue or good news. You seem "hell-bent" on pushing your "TSLA-Long" agenda outside of the investment thread, even calling people you don't know "Shorts," "Trolls" in a confrontational manner (including myself a while back and I own TSLA stock)... As others have said, I too am glad you are not a moderator here on TMC... Yes, I can get very passionate and negative on a specific issue, like the FSD ordering page "Really" matter, so I'm not proclaiming that I'm a saint; but you have a different issue going on it seems... Look in the mirror and you might see a real "troll". Later.
Last edited:
I am sure Tesla is working on HW3.5 or whatever it will be called but I think that is a separate issue. HW2/2.5 being based on nVidia products have a much higher chance of being obsoleted similar to AP1 than the HW3 FSD computer does, which supposedly is the first product Tesla really designed for the task of the FSD, and which is an in-house design.
So I think it would make sense for Tesla to try and cut HW2/2.5 loose by wrapping up EAP and moving all development to HW3+ (HW3 here including HW2/2.5 FSD computer retrofits), while mothballing HW2/2.5 development similar to AP1.
Most of the issues with Smart Summons seem to be deliberate results of not doing things a certain way.
Like the path planning seems to be a result of not using maps.
Not always trying to get in the right most part of the lane seems to be a deliberate choice versus just an accidental.
There are very few things I'm seeing in the latest videos that aren't indicative of design limitations. So I don't see it being improved much from now until release. Like I don't believe they intend for it to stop at stop lines.
Just like most of the issues with Summons or Self-park as they exist now are the result of deliberate limitations. Limitations so severe that I only use summons occasionally.
They might be able to smooth out curb detection a bit, but I think it's at a point where it will take quite a bit of work to get smart summons to the next level.
So I expect the wide release to happen within 3 weeks where its largely in its current state.
Like the path planning seems to be a result of not using maps.
Not always trying to get in the right most part of the lane seems to be a deliberate choice versus just an accidental.
There are very few things I'm seeing in the latest videos that aren't indicative of design limitations. So I don't see it being improved much from now until release. Like I don't believe they intend for it to stop at stop lines.
Just like most of the issues with Summons or Self-park as they exist now are the result of deliberate limitations. Limitations so severe that I only use summons occasionally.
They might be able to smooth out curb detection a bit, but I think it's at a point where it will take quite a bit of work to get smart summons to the next level.
So I expect the wide release to happen within 3 weeks where its largely in its current state.
Similar threads
- Article
