What will happen within the next 6 1/2 weeks?

[daniel]

I think Wikipedia got it right on the topic we were discussing. Allow me to re-post the Wikipedia quote and my brief note:

Artificial neural networks (ANN) or connectionist systems are computing systems vaguely inspired by the biological neural networks that constitute animal brains. ... The original goal of the ANN approach was to solve problems in the same way that a human brain would. ...

Note the words "vaguely inspired by," and that solving problems "the same way a human brain would" is merely "The original goal" of neural networks.

Wikipedia is not saying that neural networks operate like the brain. It's saying that the original goal was to do so.

... Tesla is one step closer to "feature complete". ...

And when I walk to the beach I am many steps closer to China. But China is effectively just as far away from me as it was. I'm not saying that "feature complete" (a useless term IMO) is as far away as China. Just that one step closer might be less than we'd like. And City NoA is one chicken I won't count until it's hatched.

 

[diplomat33]

And when I walk to the beach I am many steps closer to China. But China is effectively just as far away from me as it was. I'm not saying that "feature complete" (a useless term IMO) is as far away as China. Just that one step closer might be less than we'd like. And City NoA is one chicken I won't count until it's hatched.

Tesla is a lot closer to releasing City NOA than that.

 

[EVNow]

And when I walk to the beach I am many steps closer to China. But China is effectively just as far away from me as it was. I'm not saying that "feature complete" (a useless term IMO) is as far away as China. Just that one step closer might be less than we'd like. And City NoA is one chicken I won't count until it's hatched.

If you think its so far away, you are wasting your time here

 

[daniel]

Tesla is a lot closer to releasing City NOA than that.

I hope you are right.

If you think its so far away, you are wasting your time here

I'm here because I hope I'm wrong.

 

[diplomat33]

I hope you are right.

I think so. The new visualizations are a good indication that Tesla has made a ton of progress towards City NOA and "traffic light and stop sign response" is already in EA testing now. City NOA will definitely come out in 2020.

I'm here because I hope I'm wrong.

This guy who is in Early Access predicts that Tesla is less than 5 years away from being able to have cars self-deliver to a customer's house.

Third Row Podcast on Twitter

 

[EVNow]

I'm here because I hope I'm wrong.

Same here !

Seriously - I think it is probably too optimistic to expect City NOA dropped on all cars one day. We are going to see a feature build up that will finally give us City NOA.

BTW, anyone still thinks Tesla has a separate extremely good HW3 version they will release to get to FSD in no time ? Now that we have had HW3 cars for a while and they are slowly trying to introduce features to HW3 cars should be proof enough that there is no secret HW version.

 

[Shateam]

Tesla is giving some EA access to the traffic light response feature that actually stops the car:

3rd Row Tesla on Twitter

"So the big surprise this week is this:

Even though only a select few Early Access users will have the code that actually stops the car enabled...

I would have expected to see videos of this posted to YouTube by now if this was true...

 

[DrDabbles]

Yes, Elon missed another deadline. In other breaking news, water is wet.

Traffic light response is in EA testing now and "city NOA" will follow. And "city NOA" is the last piece of "feature complete". So Tesla is one step closer to "feature complete". Now, how long it will take to go from "feature complete" to "sleep in your car while it self-drives" is anybody's guess. Of course, Elon is being overly optimistic again. I think he is crazy to think that it will only take another year to get to "sleep in your car while it self-drives". Personally, I think it will take longer than that.

But Elon is not entirely wrong in his last statement. While Tesla did not achieve the coast to coast FSD demo this year, he is right about the second part that Tesla is not planning to do the coast to coast FSD demo with software that is not available to the public. Who knows when Tesla will have that capability but whenever they do, it will be with software that the public who purchased FSD will also have in their cars at the same time.

The problem with this line of thinking is that "feature complete" means "the thing we promised has been delivered". So, when Elon promised on multiple occasions for multiple years that your car would drive while you were asleep, or that you could summon it from across the country (US), and the actual website described the features as such, that started a bit of a problem. You can't tell a customer you delivered something when that thing doesn't actually do what was promised.

I also take serious issue with the very idea that FSD could possibly be "feature complete" if it can't do those tasks. If it can not drive itself while you sleep, if it can not drive through a city, town, or otherwise incorporated or unincorporated area without aide, then it is not self driving. This concept that the features could somehow be present but not actually work is asinine at its very core. The Tesla Apologists certainly have a lot of work ahead of them in 2020.

Now, on to an actual improvement. The FSD "sneak peak" (which isn't actually a preview of FSD, but rather image recognition) is clearly interpreting signs, painted wording, and indications on the roadway. That's a very important improvement, and presumably means their next step is interpreting the meaning of those traffic control mechanisms and using them for decision making (should I move into that other lane, can I turn left here, where do I stop for the stop sign, should I slow down for something up ahead, etc.). That's big for Tesla, and a critical step on the long hard road to autonomy.

 

[diplomat33]

The problem with this line of thinking is that "feature complete" means "the thing we promised has been delivered". So, when Elon promised on multiple occasions for multiple years that your car would drive while you were asleep, or that you could summon it from across the country (US), and the actual website described the features as such, that started a bit of a problem. You can't tell a customer you delivered something when that thing doesn't actually do what was promised.

You misunderstand what feature complete is. In software engineering, feature complete is when all the features are coded it but it still needs to go through beta testing and bug fixes before it is delivered to the customer. So feature complete does not mean the finished product delivered to the customer. In fact, a company would never deliver feature complete to a customer because the product would be riddled with bugs. Of course, with Tesla, it is different since they do deliver beta features and do the beta testing after the product is delivered to the customer. But the concept is still the same. Feature complete is not the finished product, it is just the first stage where the features are done but not perfect yet. Even with Tesla, they always do improvements after they deliver AP features to the fleet.

Tesla is not claiming that FSD feature complete is the finished FSD where you can sleep in your car. When Elon talks about sleeping in your Tesla, he was never talking about "feature complete". He was talking about the final product after "feature complete".In fact, during the earnings call, he explained the 3 stages of FSD development:
Stage 1: feature complete: the car has the essential features to self-drive from A to B but requires driver supervision and driver intervention.
Stage 2: Tesla refines the software until they are confident that driver supervision is not needed anymore.
Stage 3: Tesla may refine the software some more until regulators agree that driver supervision is not needed anymore.

So "sleeping in your Tesla while it drives you" will come at stage 2 or 3, well after "feature complete".

 

[diplomat33]

I would have expected to see videos of this posted to YouTube by now if this was true...

EA is under a NDA where they are not allowed to share information. Although, apparently from one source I read, Elon does sometimes allow an exception or two and allow leaks like with Smart Summon. But in most cases, EA is under strict rules that some things have to be kept secret.

 

[am_dmd]

When Musk let's anybody put his kiddos in a FSD Tesla I will be the first one to pay for whatever Tesla wants to charge.

 

[_jal_]

I leased my first MS and bought EAP. When I turned it in in October, NoA still didn’t work. As the person responsible for being in control of the car and taking over at any time, I judged the feature to be dangerous and not worth the risk of engaging. I tried it, but had to stop using it.

I guess it was feature complete, but I never got what I paid for. That’s why I don’t care if FSD is feature complete. It’s almost just an internal engineering benchmark. I will buy FSD when my car can fully drive itself - a pretty self explanatory term implied by a feature called “Full Self Driving”

 

[DrDabbles]

You misunderstand what feature complete is.

I love it when people try to explain my industry to me.

In software engineering, feature complete is when all the features are coded it but it still needs to go through beta testing and bug fixes before it is delivered to the customer.

There are a couple pretty important facts here. First, this isn't what Elon is describing now. Second, if you're an independent developer you do not consider this delivered software, so the project isn't complete. Development work by definition isn't complete. You are not getting paid for this milestone.

So feature complete does not mean the finished product delivered to the customer.

Lovely, then Tesla shouldn't recognize the revenue and they shouldn't claim they've completed FSD.

In fact, a company would never deliver feature complete to a customer because the product would be riddled with bugs.

If your code is "feature" complete and "riddled with bugs", you're probably going on a performance improvement program. Partially because you clearly need TDD and partly because you've somehow created features without considering what it needs to do. If I asked you to develop a billing system and it's "riddled with bugs", then what the hell have you been doing exactly?

Of course, with Tesla, it is different since they do deliver beta features and do the beta testing after the product is delivered to the customer.

Nothing like asking untrained members of the public to experiment with a deadly machine, right?

Feature complete is not the finished product, it is just the first stage where the features are done but not perfect yet.

Good, then Tesla shouldn't be pretending that this is a meaningful event for customers.

Tesla is not claiming that FSD feature complete is the finished FSD where you can sleep in your car.

Well, Elon isn't pretending this anymore. He was pretending that fall asleep autonomy would be delivered by end of 2019. Which is now 4 days away. And it looks like the "none" option on the survey for this thread wins it.

When Elon talks about sleeping in your Tesla, he was never talking about "feature complete".

I get that you're a true believer, and that's fine. But you've engaged in the largest goalpost moving I've ever seen in my life. Elon's statement was that you'd have a self driving autonomous car by end of 2019. Robotaxi would begin at the start of 2020 as a service people could use. That claim was this year. Now he has an army of fans toeing his new line that they'll be delivering "feature complete" FSD by end of 2019. Which they also absolutely will not be doing.

Something you don't seem to understand is that the features are not complete. If the vehicle can not drive itself at all in any condition, then it's not feature complete or complete in any other way. You can break internal project deliverables down however you like, and perhaps you should become an AGILE certified scrum leader at Tesla. But the fact is that none of this is delivered to customers, and none of this means the car is self driving. The feature from the customer's perspective- the user story- is "I want my car to drive itself". It does not do that, so the feature is not complete. If you're a software developer and you think delivering a broken feature is the same as delivering a feature to a customer, you should look for ways to improve your craft or perhaps find a different industry to join. May I suggest Presales Engineer.

 

[diplomat33]

Good, then Tesla shouldn't be pretending that this is a meaningful event for customers.

"Feature complete" will be a meaningful milestone because it will be a big jump in AP capabilities.

Well, Elon isn't pretending this anymore. He was pretending that fall asleep autonomy would be delivered by end of 2019. Which is now 4 days away. And it looks like the "none" option on the survey for this thread wins it.

I get that you're a true believer, and that's fine. But you've engaged in the largest goalpost moving I've ever seen in my life. Elon's statement was that you'd have a self driving autonomous car by end of 2019. Robotaxi would begin at the start of 2020 as a service people could use. That claim was this year. Now he has an army of fans toeing his new line that they'll be delivering "feature complete" FSD by end of 2019. Which they also absolutely will not be doing.

That's not completely accurate. Yes, Elon said robotaxis by 2020 but Elon said "fall asleep autonomy" would be end of 2020, not 2019. And he said "feature complete" by end of 2019.

Here is what Elon said this year:

“I think we will be feature-complete, full self-driving this year, meaning the car will be able to find you in a parking lot, pick you up, take you all the way to your destination without intervention this year. I would say I am certain of that, that is not a question mark. However, sometimes people will extrapolate that to mean now it works with 100 per cent certainty, requiring no observation, perfectly – this is not the case.”

"My guess as to when we would think it’s safe for somebody to essentially fall asleep and wake up at the destination? Probably towards the end of next year."

Something you don't seem to understand is that the features are not complete. If the vehicle can not drive itself at all in any condition, then it's not feature complete or complete in any other way. The feature from the customer's perspective- the user story- is "I want my car to drive itself". It does not do that, so the feature is not complete. If you're a software developer and you think delivering a broken feature is the same as delivering a feature to a customer, you should look for ways to improve your craft or perhaps find a different industry to join. May I suggest Presales Engineer.

Feature complete is a full self-driving car, just not 100% perfect, thus requiring driver supervision at first.

Again, this is how Elon described "feature complete":
“I think we will be feature-complete, full self-driving this year, meaning the car will be able to find you in a parking lot, pick you up, take you all the way to your destination without intervention this year. I would say I am certain of that, that is not a question mark. However, sometimes people will extrapolate that to mean now it works with 100 per cent certainty, requiring no observation, perfectly – this is not the case.”

 

[daniel]

... I don’t care if FSD is feature complete. It’s almost just an internal engineering benchmark. ...

Strike "almost" from the above. "Feature complete" is exactly an internal engineering benchmark and means nothing to drivers, though it might be interesting to fans of the process who enjoy the story of how an engineering project progresses.

Nothing like asking untrained members of the public to experiment with a deadly machine, right?

I don't consider AP/EAP to be unfinished software sent out to consumers to debug. Used properly, I believe it makes me a safer driver. It has been improved since I bought my car, but from the start it was a significant safety feature. The caveat is that you have to use it properly, which is to say, stay alert and ready to take over, and only engage it when it can do a better job than you can. I believe that Tesla calls it "beta" not because it truly is beta software, but because having chosen to call it "autopilot," which it is not, they need to very strongly impress upon people that it is not autopilot, but merely a set of driver-assist features.

Cars are deadly machines, and humans are poor operators of them. AP/EAP has made them a little bit safer. It seems to have hit a plateau, and I am not expecting significant improvement in the next few years, but I do expect improvements to come.

The above is my opinion based on my own experience.

 

[DrDabbles]

"Feature complete" will be a meaningful milestone because it will be a big jump in AP capabilities.

A big jump in AP capabilities that nobody can use. Which makes them not capabilities.

That's not completely accurate. Yes, Elon said robotaxis by 2020 but Elon said "fall asleep autonomy" would be end of 2020, not 2019. And he said "feature complete" by end of 2019.

Well, he said that after he said the cars would totally drive themselves and that you could summon them from across the country. Like I said, the goalposts are moving here.

Feature complete is a full self-driving car, just not 100% perfect, thus requiring driver supervision at first.

A full self driving car implies it's fully self driving. If it requires driver supervision it's not fully self driving, it's partially self driving. We've been over that a millions times, so let's accept that you think Level 3 qualifies as self driving and I accept that Level 4 mostly qualifies and Level 5 is actually fully self driving.

I don't consider AP/EAP to be unfinished software sent out to consumers to debug.

Certainly. And those features do what they claim in the tin. They're driver assistance functions, and they assist us. They fail in the exact ways that Tesla predicted on their blog back in 2016. Those functions are specifically not advertised as self driving.

Cars are deadly machines, and humans are poor operators of them. AP/EAP has made them a little bit safer. It seems to have hit a plateau, and I am not expecting significant improvement in the next few years, but I do expect improvements to come.

Attentive humans are amazing operators of vehicles, and humans operating vehicles within the limits of the vehicle's and the operator's capabilities are supreme operators. The problem is that humans stop paying attention, and study after study is concluding that humans relying on ADAS systems to drive for them is less safe than a human paying attention because that system lulls them into trusting it and the human is more apt to stop paying attention.

AP/EAP is nowhere near as good at operating a vehicle as an attentive human. Not even close. The argument you should be making is that an attentive human plus an ADAS system is a superior solution which will hopefully only get better. But if you just stop paying attention, right now AP/EAP will crash into things.

 

[daniel]

AP/EAP is nowhere near as good at operating a vehicle as an attentive human. Not even close. The argument you should be making is that an attentive human plus an ADAS system is a superior solution which will hopefully only get better. But if you just stop paying attention, right now AP/EAP will crash into things.

This is exactly what I've been saying: AP/EAP with an attentive driver at the wheel is safer than a human driver alone.

 

[DrDabbles]

This is exactly what I've been saying: AP/EAP with an attentive driver at the wheel is safer than a human driver alone.

You and I have a history of agreement here. You're dangerous and subversive and your brain needs more washing.

Off topic, how's it owning a Tesla on the Islands? Maui doesn't have a supercharger, right? Did you have to get the car shipped from Oahu, or did it come from the mainland? Do you live in the more jungle areas or the more arid areas of the island, and do you notice the HVAC smell if you live in the rainy areas?

 

[Alset Srotom]

EA is under a NDA where they are not allowed to share information. Although, apparently from one source I read, Elon does sometimes allow an exception or two and allow leaks like with Smart Summon. But in most cases, EA is under strict rules that some things have to be kept secret.

Im not sure where this whole signed NDA thing came from, but as someone in early access, I can tell you it's not true.

I can also tell you that no regular person in the early access program has traffic light/stop sign response...

 

[diplomat33]

A full self driving car implies it's fully self driving. If it requires driver supervision it's not fully self driving, it's partially self driving. We've been over that a millions times, so let's accept that you think Level 3 qualifies as self driving and I accept that Level 4 mostly qualifies and Level 5 is actually fully self driving.

It's not a matter what your or I want to think. The SAE defines L3, L4 and L5 as what is colloquially referred to as "full self-driving" because they can perform the entire DDT when they are on. So L3, L4 and L5 are all "full self-driving".