Elon: "Feature complete for full self driving this year"

[mongo]

@mongo All you say is true of course. It does take a certain amount of faith though — given Tesla’s past lapses in delivering what they say they would — to think HW3 will magically unleash the data-driven edge.

I do think Tesla has a better-trained NN that will benefit from HW3, no doubt, but again that in itself is no proof of data-driven edge until Tesla is really showing something better than others tech-wise...

As said they do have the deployment advantage.

Oh I get the high level of wait and see. No argument there. Only pointing out that, even with the first HW3 units in the wild, nothing has been made public that lets one make an evaluation of the state of Tesla's internal NN development or data gathering. And with the fleet size, if HW3 has the extra bandwidth to run the current NN while collected new/ different data (which it should), Tesla can overcome a lot of source data related technical debt, if it needs to.

The post would have made more sense if it were a reply to one that said "OMG HW3 is out!!!!! Telsa's NN is so boss. Squee!!!!"

 

[electronblue]

The post would have made more sense if it were a reply to one that said "OMG HW3 is out!!!!! Telsa's NN is so boss. Squee!!!!"

Well. That reply would not have started with ”All you say is true of course.”

 

[mongo]

Well. That reply would not have started with ”All you say is true of course.”

Well yah! (Just so we're clear, I was talking about bladers post, not yours. )

 

[diplomat33]

Every single problems in AP1 exist in AP2 and some is actually worse. Phantom braking = Worse, Diving for Dividers = Worse (infact did not exist in AP1 from what i can tell), Running into parked/stopped cars = Worse/Same, Running into stationary objects = Worse/Same.

Totally 1000% False. I drive a AP2.5 every single day and I can categorically say that the issues are not worse than AP1. In fact, it's gotten better. I've had less phantom braking than before, no phantom braking in the last 4 months. I've never seen my car "dive for the dividers". I use AP on city streets and it does not try to run into parked, stopped cars either. And when I come up on a stopped car at a red light, AP does stop.

 

[diplomat33]

Because after almost 3 years of "billions of miles of data" according to the faithful.
Situations such as this is still common: Jay in Shanghai on Twitter

I did not see anything in that video that AP was activated. So how do we know it was AP's fault and not driver fault? I can tell you that I've been in several instances with a stopped car at a red light in front of me and every single time, EAP has stopped for me.

 

[Az_Rael]

Totally 1000% False. I drive a AP2.5 every single day and I can categorically say that the issues are not worse than AP1. In fact, it's gotten better. I've had less phantom braking than before, no phantom braking in the last 4 months. I've never seen my car "dive for the dividers". I use AP on city streets and it does not try to run into parked, stopped cars either. And when I come up on a stopped car at a red light, AP does stop.

You must not be on 2019.5.15. The phantom braking in that release is horrible. Way worse than prior releases. I can't wait to get off it and am sorry I accepted the firmware update for the first time in a long time.

Our AP1 car does not exhibit that level of phantom braking.

 

[diplomat33]

You must not be on 2019.5.15. The phantom braking in that release is horrible. Way worse than prior releases. I can't wait to get off it and am sorry I accepted the firmware update for the first time in a long time.

Our AP1 car does not exhibit that level of phantom braking.

No, I have 2019.5.15. No phantom braking for me.

 

[Az_Rael]

No, I have 2019.5.15. No phantom braking for me.

Consider yourself lucky then.

My car has been goosing the brakes at freeway speeds a lot in that release. It appears to be locking onto slower moving cars in adjacent lanes (straight roads, not curves) and briefly thinking the other car is encroaching when it is not. I have to hover my foot over the accelerator when passing slower cars now, it is that frequent.

 

[diplomat33]

Consider yourself lucky then.

My car has been goosing the brakes at freeway speeds a lot in that release. It appears to be locking onto slower moving cars in adjacent lanes (straight roads, not curves) and briefly thinking the other car is encroaching when it is not. I have to hover my foot over the accelerator when passing slower cars now, it is that frequent.

Sorry to hear that. The only issue I have with 2019.5 is that TACC does not appear to be as smooth as before. Hopefully, the next update will fix your phantom braking issues.

 

[Bladerskb]

Totally 1000% False. I drive a AP2.5 every single day and I can categorically say that the issues are not worse than AP1. In fact, it's gotten better. I've had less phantom braking than before, no phantom braking in the last 4 months. I've never seen my car "dive for the dividers". I use AP on city streets and it does not try to run into parked, stopped cars either. And when I come up on a stopped car at a red light, AP does stop.

I'm talking about overall experience not the experience of one man in a single location and the posts above proves my point without even trying.

 

[diplomat33]

I'm talking about overall experience not the experience of one man in a single location and the posts above proves my point without even trying.

No, you are picking 1 example where the car hits a truck or 1 example where the car "dives for a divider" and claiming that these examples represent an overall experience. These are not common issues for all users.

 

[Bladerskb]

No, you are picking 1 example where the car hits a truck or 1 example where the car "dives for a divider" and claiming that these examples represent an overall experience. These are not common issues for all users.

No i didn't, I only posted one example. There are dozens of videos of AP2 diving for dividers, dozens. I'm sorry but you are not even qualified to have this discussion. I have been here longer than you and have been involved in the discussion of autonomous systems and Tesla's AP way longer than you. Its not even close. Stop putting your experience over everyone's else experience as a whole.

 

[diplomat33]

I'm sorry but you are not even qualified to have this discussion.

I am not taking away from your knowledge about self-driving but to tell a Tesla owner like me that he is not even qualified to discuss the issue is absolutely ludicrous! As a Tesla owner, I am absolutely qualified to discuss how AP works because I use it everyday.

 

[Bladerskb]

I am not taking away from your knowledge about self-driving but to tell a Tesla owner like me that he is not even qualified to discuss the issue is absolutely ludicrous! As a Tesla owner, I am absolutely qualified to discuss how AP works because I use it everyday.

Because you weren't here to experience the Pre-AP1 issues/complaints, the AP2+ transitional issues/complaints from people who had AP1 and the aftermath AP2+ issues/complaints/comparison from people who had AP1. You weren't have for the minute by minute, hour by hour, day by day discussion and debates. Therefore you are incapable of discussing issues that AP1 had, their frequency and how it compares to AP2 because you weren't here for that. If you were you would know that phantom braking was the biggest issues of AP2 from AP1 owners. But i don't blame you. Its similar to people who now vow and believe without reasonable doubt that Tesla will have FSD Level 5 by end of 2019/2020. They weren't here when all the people who now oppose them were saying the same thing only with a 2018 date. You're a second gen tesla fan, Its not your fault. But don't act like you are well informed of the struggles of fans that came before. Do you think these fans were Angry back in 2016/2017? no they were just as defiant as you are now when anyone questioned Tesla.

 

[diplomat33]

Because you weren't here to experience the Pre-AP1 issues/complaints, the AP2+ transitional issues/complaints from people who had AP1 and the aftermath AP2+ issues/complaints/comparison from people who had AP1. You weren't have for the minute by minute, hour by hour, day by day discussion and debates. Therefore you are incapable of discussing issues that AP1 had, their frequency and how it compares to AP2 because you weren't here for that. If you were you would know that phantom braking was the biggest issues of AP2 from AP1 owners. But i don't blame you. Its similar to people who now vow and believe without reasonable doubt that Tesla will have FSD Level 5 by end of 2019/2020. They weren't here when all the people who now oppose them were saying the same thing only with a 2018 date. You're a second gen tesla fan, Its not your fault. But don't act like you are well informed of the struggles of fans that came before. Do you think these fans were Angry back in 2016/2017? no they were just as defiant as you are now when anyone questioned Tesla.

Did you experience AP1 issues? Did you own a pre-AP or AP1 Tesla? Please tell me. I thought you did not own a Tesla. And you only joined this forum in 2016. You have no authority greater than me to discuss AP issues.

And, I may not have direct experience with AP1 but I can certainly share my experiences with AP2.5.

Spare me the condescending attitude please!

 

[electronblue]

And, I may not have direct experience with AP1 but I can certainly share my experiences with AP2.5.

Your message did come across as you saying AP2.5 was better than AP1 in those points, so maybe make that clearer next time.

 

[diplomat33]

Your message did come across as you saying AP2.5 was better than AP1 in those points, so maybe make that clearer next time.

Sure. I am comparing my direct experience with AP2.5 to what I have read and seen about AP1. So it is not a direct 1-1 comparison I will admit. But Bladder made a categorical statement that AP2 was basically total crap in a bunch of issues. I feel like I can certainly share my input that my direct experience does not match that assessment.

Frankly, I just get very defensive because I have a great experience with AP2.5 so I don't appreciate a non-Tesla owner like bladder telling me that I should just shut up and accept that AP2.5 is crap.

 

[electronblue]

Sure. I am comparing my direct experience with AP2.5 to what I have read and seen about AP1. So it is not a direct 1-1 comparison I will admit. But Bladder made a categorical statement that AP2 was basically total crap in a bunch of issues. I feel like I can certainly share my input that my direct experience does not match that assessment.

Having driven both, AP1 at its finest was actually really good — too good for the inherent limitations that it had, namely the lack of cross-traffic support in EyeQ3 and no redundant system like Lidar to compensate, and only a narrow camera given to it that didn’t allow for the tightest of bends or sharpest of hills due to the camera losing sight... other than that you could basically use it hands-free for miles and miles, with far higher reliability than AP2 has shown so far.

Also one thing AP1 did really well was follow a lead car. It could basically follow a lead car without road markings at all...

But then AP1 is a shadow of its former self too, because of the later addition of a poor attempt to use radar instead of Lidar for cross-traffic detection — that is what brought the dreaded ghost braking events... and of course Tesla introduced the frequent nags for which there is no remedy in a car without steering wheel touch/camera sensors, so it doesn’t always detect you are holding the steering wheel...

Boy, was AP1 good at its best — and that is pretty much a testament to how good the EyeQ3 vision engine was/is. Also, it read speed-limit signs (already EyeQ2 in 2010 did that by the way, read those small extra signs underneath the speed limits too! in 2010!). AP2+ does not read speed-limit signs yet.

 

[diplomat33]

Having driven both, AP1 at its finest was actually really good — too good for the inherent limitations that it had, namely the lack of cross-traffic support in EyeQ3 and no redundant system like Lidar to compensate, and only a narrow camera given to it that didn’t allow for the tightest of bends or sharpest of hills due to the camera losing sight... other than that you could basically use it hands-free for miles and miles, with far higher reliability than AP2 has shown so far.

Also one thing AP1 did really well was follow a lead car. It could basically follow a lead car without road markings at all...

But then AP1 is a shadow of its former self too, because of the later addition of a poor attempt to use radar instead of Lidar for cross-traffic detection — that is what brought the dreaded ghost braking events... and of course Tesla introduced the frequent nags for which there is no remedy in a car without steering wheel touch/camera sensors, so it doesn’t always detect you are holding the steering wheel...

Boy, was AP1 good at its best — and that is pretty much a testament to how good the EyeQ3 vision engine was. Also, it read speed-limit signs (already EyeQ2 in 2010 did that by the way, read those small extra signs underneath the speed limits too! in 2010!). AP2+ does not read speed-limit signs yet.

Thanks. I know from several reports that AP1 was excellent. I am not knocking AP1. I am just disagreeing that AP2.5 is "way worse" especially considering that my experience with AP2.5 is also quite excellent.

 

[Bladerskb]

so I don't appreciate a non-Tesla owner like bladder telling me that I should just shut up and accept that AP2.5 is crap.

I never said that..

In other news, I'm still waiting on any evidence that the last ~3 years of billions of miles of data has infact given Tesla 3+ years of immense lead in self driving as have been claimed for years and continue to be claimed.

I guess i can wait while i ponder on the below.

Tesla Autopilot Fender Bender - Streamable