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

[drkeys]

Yup real substance right there, got your model 3, now you're an AI/character judgement expert that posts MEME's and backing up your fanboy comments

 

[diplomat33]

Yup real substance right there, got your model 3, now you're an AI/character judgement expert that posts MEME's and backing up your fanboy comments

George Hotz is a hack. All he's done is a write some code to create a poor man's version of AP1. Karpathy has created neural nets that make Nav on AP possible. To say that he is better than Karpathy is laughable. And comma.ai raised $5 million. Tesla is a billion dollar company. That is why I posted the meme.

 

[drkeys]

yeah because he wants to retrofit autopilot and assisted driving for the masses. for Honda's, Toyotas, cars that people use everyday. I'm sure if he has the resources from Tesla's billion dollar company, he can make neural net and FSD work a lot better and release a lot faster than Karpathy could. Because he himself took on MobileEye and Tesla AP1 and make that available for regular cars and make sure they're compatible.. that is a major accomplishment with bare minimum resources (how long did it took Tesla to get autopilot?)

 

[diplomat33]

yeah because he wants to retrofit autopilot and assisted driving for the masses. for Honda's, Toyotas, cars that people use everyday. I'm sure if he has the resources from Tesla's billion dollar company, he can make neural net and FSD work a lot better and release a lot faster than Karpathy could. Because he himself took on MobileEye and Tesla AP1 and make that available for regular cars and make sure they're compatible.. that is a major accomplishment

It's an accomplishment sure but not a major one. You can't compare retrofitting a very basic driver assist with what Tesla, Mobileye and others are doing. Hotz is cobbling together a L2 driver assist to old cars. So yeah, he's making L0 and L1 cars into L2 cars. Tesla, Mobileye are working on L4 autonomy. Sorry. Not in the same ballpark. And no, I don't believe Hotz could do Karpathy's job better if he had Tesla's resources.

 

[Tjhappel]

He’s definitely had the ability to raise money with his early fame and for whatever reason he hasn’t so....

Not that I know anything but I am a huge fan of both hotz and Elon but one of them is making a huge product and the other has a small product that raised 5 million. Not saying hotz isn’t smarter in the self driving space (likely he is) but Elon seems to be getting further ahead.

 

[verygreen]

It's an accomplishment sure but not a major one. You can't compare retrofitting a very basic driver assist with what Tesla, Mobileye and others are doing. Hotz is cobbling together a L2 driver assist to old cars. So yeah, he's making L0 and L1 cars into L2 cars. Tesla, Mobileye are working on L4 autonomy. Sorry. Not in the same ballpark. And no, I don't believe Hotz could do Karpathy's job better if he had Tesla's resources.

Have you driven one?

I was in a Model S with comma.ai box retrofitted (pre-AP car) and I was kind of shocked how good it was (granted we should not say it was done by geohot singlehandedly, they have a bunch of devs over there at comma.ai). It displays the same things AP1 (and at times AP2) would on the IC and otherwise looked very native.

It took Tesla 2+ years to reach this level with a lot more resources and they only do it for three cars they make themselves and intimately know where as comma.ai works on a whole bunch of makes/models often based on reverse-engineered data.

 

[Tjhappel]

Yes we all wish hotz would have taken the Tesla job Elon offered.... it would or could have been amazing but it didn’t happen.

It’ll be interesting to see v3. Can’t wait till very green gets his hands on it.

Thanks for all the info, it’s fun to read and speculate on.

 

[ord3r]

Hotz is godlike, but I’m also sure he’s the reason Elon thought autonomy was “solved” in 2016 and that we are living in a simulation.

We would be seeing some insane stuff if he was at Tesla. Would have been an interesting timeline to say the least.

 

[electronblue]

George Hotz is a hack. All he's done is a write some code to create a poor man's version of AP1. Karpathy has created neural nets that make Nav on AP possible. To say that he is better than Karpathy is laughable. And comma.ai raised $5 million. Tesla is a billion dollar company. That is why I posted the meme.

Comma.ai may not be positioned to change the world but so far they actually seem to have a more reliable vision engine than Tesla in production. That is not nothing given the comparative lack of resources at Comma.ai.

I agree Hotz had a big mouth in 2016 just like Musk. Hotz has toned down things a bit since. Musk has doubled down. That is not necessarily indicative of progress though. It may also mean one is learning...

Musk needed Hotz back in 2016 far more than we realized at the time.

 

[CarlK]

Besides doing his Ph.D and few internships, and finally got a first gig job at Tesla, he really has not accomplished anything except did some research paper

He got the PhD from the university with one of the top AI program in the world. He helped design and taught the popular course "Convolutional Neural Networks for Visual Recognition" in Standford's Vison and Learning lab. He was a founding scientist of OpenAI and also had interned at Google's Deep Mind project while he's a student. Pretty busy few years if that's what you meant. If his young age bothers you Einstein once said "A person who has not made his great contribution to science before the age of thirty will never do so.".

Hold on folks, the pace of development is now "exponential"!

It's actually true. That's the nature of machine learning and its biggest potential threat to humanity. It'll go from sub-human to super-human so fast we will not even have time to notice it. Contrary to what many may believe there is nothing so special of human intelligence scale.

While the interview mostly lacked detail (but high level of confidence, must have pissed Blader off! lol), there was again validation that Tesla uses data from disengagements of Autopilot to send back to improve edge case learning. It sounded like only in specific instances though.

Made me think, I think there is going to be rapid iteration on what events are triggered to send back based on the current state of the FSD algorithm. For instance they only probably started collected some triggered disengagements of lane changes when autopilot suggested it, and the user declined (probably not accurate enough) or when the user accepted, but then disengaged (realizing some object /problem).

Meanwhile, stuff like traffic light detection has been developed with in-house data, but once it's deployed to a certain level of accuracy, only then will certain events that the user disagrees with the algorithm will a trigger save the data and send the edge case back to Tesla.

This means that Tesla's incremental development / deployment cycle is not just about making money / providing a nice L2/L3 product, but it also directly is beneficial to iteratively collecting edge cases over time as the algorithm is in good enough shape to trust the algorithm estimate / user input divergence.

It will probably applicable to all cases. In his May 2018 presentation Andrej Karpathy specifically mentioned edge cases in traffic lights and road signs is the reason they need more data to cover them.

Hotz is godlike, but I’m also sure he’s the reason Elon thought autonomy was “solved” in 2016 and that we are living in a simulation.

We would be seeing some insane stuff if he was at Tesla. Would have been an interesting timeline to say the least.

Not true. Elon's involvement in AI went way back than that.

Hotz's did not like anyone else's except he still had a few good things to say about Tesla's program. He particularly disliked Mobileye's technical approach and public stance. Pretty obvious Elon already had his project in place already at the time when he's trying to recruite Hotz.
'Jokers' and 'sellouts': The cockiest guy in self-driving cars just roasted his competition

BTW you see those Tesla haters praising Hotz but it would be a totally different story if he indeed had come to work for Tesla. Hmmm they might change the tone now that they found he actually liked Tesla.

 

[CarlK]

Yup real substance right there, got your model 3, now you're an AI/character judgement expert that posts MEME's and backing up your fanboy comments

I was actually wanting to ask you this same question. You were trying to belittle Kaparthy what is your background in AI or what Tesla cars you own?

 

[Bladerskb]

Going back to Elon's statement of "Feature complete for full self driving this year".
Elon now says in response to the questions.

Question on how long will Tesla still require human supervision

Tesla's full self driving as still for a time to come? The requiring supervision the human being so its capabilities are powerful enough to drive but nevertheless requires a human to still be supervising just like a safety driver is in other fully autonomous vehicles?

Elon Musk: "I think it will require detecting hands on wheel for at least six months or something like that, from here really is question of like, from regulatory standpoint, what how much safer than a person does autopilot need to be for it took to be okay to not monitor the car."

Question on adding Driver Monitoring System

Elon Musk: "I think this was all really going to be swept, I mean, the system is improving so much, so fast, that this is going to be a moot point very soon. No, in fact, I think it will become very, very quickly, maybe and towards the end this year, but I say, I'd be shocked if not next year, at the latest that having the purse, having human intervene will decrease safety. DECREASE!"

Conclusion:

1) Tesla will remove hands on wheel human supervision requirement in at-least 6 months (Oct 2019)
2) Tesla's "FSD" will be so good by end of this year or would be shocking if not by next year to the point that human supervision won't be required because human intervention will decrease safety.

Basically Elon is saying we will have Level 5 full autonomy in 6 months.

 

[iTech]

Going back to Elon's statement of "Feature complete for full self driving this year".
Elon now says in response to the questions.

Question on how long will Tesla still require human supervision



Elon Musk: "I think it will require detecting hands on wheel for at least six months or something like that, from here really is question of like, from regulatory standpoint, what how much safer than a person does autopilot need to be for it took to be okay to not monitor the car."

Question on adding Driver Monitoring System

Elon Musk: "I think this was all really going to be swept, I mean, the system is improving so much, so fast, that this is going to be a moot point very soon. No, in fact, I think it will become very, very quickly, maybe and towards the end this year, but I say, I'd be shocked if not next year, at the latest that having the purse, having human intervene will decrease safety. DECREASE!"

Conclusion:

1) Tesla will remove hands on wheel human supervision requirement in at-least 6 months (Oct 2019)
2) Tesla's "FSD" will be so good by end of this year or would be shocking if not by next year to the point that human supervision won't be required because human intervention will decrease safety.

Basically Elon is saying we will have Level 5 full autonomy in 6 months.

From my garage down the driveway sure.

 

[diplomat33]

Going back to Elon's statement of "Feature complete for full self driving this year".
Elon now says in response to the questions.

Question on how long will Tesla still require human supervision

Elon Musk: "I think it will require detecting hands on wheel for at least six months or something like that, from here really is question of like, from regulatory standpoint, what how much safer than a person does autopilot need to be for it took to be okay to not monitor the car."

Question on adding Driver Monitoring System

Elon Musk: "I think this was all really going to be swept, I mean, the system is improving so much, so fast, that this is going to be a moot point very soon. No, in fact, I think it will become very, very quickly, maybe and towards the end this year, but I say, I'd be shocked if not next year, at the latest that having the purse, having human intervene will decrease safety. DECREASE!"

Conclusion:

1) Tesla will remove hands on wheel human supervision requirement in at-least 6 months (Oct 2019)
2) Tesla's "FSD" will be so good by end of this year or would be shocking if not by next year to the point that human supervision won't be required because human intervention will decrease safety.

Basically Elon is saying we will have Level 5 full autonomy in 6 months.

Nope. Your conclusion is wrong.

I think you are taking Musk out of context. Look at the question again. (17:14 mark in the video):

Q: Let's talk about full autonomy, if we look at the current L4 vehicles being tested now like Waymo and so on, they are only technically autonomous. They are still L2 systems with just a different design philosophy. There is always a safety driver in almost all cases and they are monitoring the system. Do you see Tesla's FSD as still for a time to come requiring supervision by the human being. So it's capabilities are powerful enough to drive but nevertheless still requires a human to supervise just like a safety driver is in another fully autonomous vehicle?

Musk: "I think it will require detecting hands on wheel for at least six months or something like that, from here really is question of like, from regulatory standpoint, what how much safer than a person does autopilot need to be for it took to be okay to not monitor the car."

So the question was clearly asking about FSD, when would Tesla's FSD no longer require driver supervision. Tesla has not released FSD yet. So clearly, the 6 months refers to 6 months after FSD is released, not 6 months from today. First, Tesla has to release "feature complete" FSD then the 6 months or so of validation begin, then the hands on wheel is removed when the validation is done.

Now, I personally think the FSD validation period will take much longer than 6 months. But that is another issue.

But no, Musk was not promising L5 in 6 months.

 

[diplomat33]

AP3 upgrades for Model 3 owners with FSD option starting "in a few months" and the price of FSD is going up after May 1. Coupled with the April 22 event, the promise of "feature complete this year", the references to the Tesla Network, and the expansion of the FSD early access program, we seem to be getting a lot of signs that Tesla is serious about FSD.

I think Tesla will push out a "FSD" software update some time this year.

 

[Bladerskb]

Nope. Your conclusion is wrong.

I think you are taking Musk out of context. Look at the question again. (17:14 mark in the video):

Q: Let's talk about full autonomy, if we look at the current L4 vehicles being tested now like Waymo and so on, they are only technically autonomous. They are still L2 systems with just a different design philosophy. There is always a safety driver in almost all cases and they are monitoring the system. Do you see Tesla's FSD as still for a time to come requiring supervision by the human being. So it's capabilities are powerful enough to drive but nevertheless still requires a human to supervise just like a safety driver is in another fully autonomous vehicle?

Musk: "I think it will require detecting hands on wheel for at least six months or something like that, from here really is question of like, from regulatory standpoint, what how much safer than a person does autopilot need to be for it took to be okay to not monitor the car."

So the question was clearly asking about FSD, when would Tesla's FSD no longer require driver supervision. Tesla has not released FSD yet. So clearly, the 6 months refers to 6 months after FSD is released, not 6 months from today. First, Tesla has to release "feature complete" FSD then the 6 months or so of validation begin, then the hands on wheel is removed when the validation is done.

Now, I personally think the FSD validation period will take much longer than 6 months. But that is another issue.

But no, Musk was not promising L5 in 6 months.

No hes talking about today. You are the one twisting things yet again. You do this everytime Elon says something ridiculous.

He even double downed by saying towards the end of the year human supervision and human intervention would actually decrease safety in response to whether he has thought about adding driver monitoring system.

Will you ever not twist things Elon say to absolve him of any responsibility? He never put a qualifier on "after FSD releases" or after "city FSD is feature complete" and his other statement on "towards the end of the year" doesnt mesh with "6 months after FSD is released"

Remember FSD is already released and is already feature complete on the highway. This is from your own mouth and Elon and Tesla who classifies NOA as FSD. So you have to contradict yourself AGAIN just to absolve elon. Which wont be new.

 

[diplomat33]

No hes talking about today. You are the one twisting things yet again. You do this everytime Elon says something ridiculous.

He even double downed by saying towards the end of the year human supervision and human intervention would actually decrease safety in response to whether he has thought about adding driver monitoring system.

Will you ever not twist things Elon say to absolve him of any responsibility? He never put a qualifier on "after FSD releases" or after "city FSD is feature complete" and his other statement on "towards the end of the year" doesnt mesh with "6 months after FSD is released"

I showed you the question and the answer to show you the full context. Musk was talking about FSD. He was not talking about today. It makes no sense that he would be talking about 6 months from today when they have not even released the rest of FSD features yet. I think you are the one twisting Elon's words to suit your narrative.

Now, Tesla might remove the nags on NOA in 6 months but they can't remove all nags from all EAP in 6 months when they have not released all the FSD features yet.

Remember FSD is already released and is already feature complete on the highway. This is from your own mouth and Elon and Tesla who classifies NOA as FSD. So you have to contradict yourself AGAIN just to absolve elon. Which wont be new.

Yes NOA is FSD according to Tesla's new list. But FSD is not feature complete yet since it still lacks the city driving part.

 

[CarlK]

AP3 upgrades for Model 3 owners with FSD option starting "in a few months" and the price of FSD is going up after May 1. Coupled with the April 22 event, the promise of "feature complete this year", the references to the Tesla Network, and the expansion of the FSD early access program, we seem to be getting a lot of signs that Tesla is serious about FSD.

I think Tesla will push out a "FSD" software update some time this year.

View attachment 396544

This Elon's tweet answered the question why he said new cars with HW3 will be an appreciating property. When autonomous sharable car becomes a reality it will worth a lot more than what cars are today. Tesla certainly will charge the right price for new cars accodingly by then. Like most anything the exact timeline is anybody's guess but it's pretty clear what Tesla's long term plan is.

 

[verygreen]

1) Tesla will remove hands on wheel human supervision requirement in at-least 6 months (Oct 2019)

I cannot beieve you chose to read his statement in the most optimistic way possible.

What he is really saying: "Tesla will most definitely NOT remove hands on requirement until Oct 2019 AND for unknown time after that"

 

[Bladerskb]

I showed you the question and the answer to show you the full context.

Where in that context did Elon say "when city driving is released and FSD city driving is feature complete"?

Musk was talking about FSD. He was not talking about today.

They already have FSD and you said it yourself literally acouple pages ago that "there's no point addressing NOA as EAP when Tesla and Elon is now calling it FSD". We literally just had this debate....

It makes no sense that he would be talking about 6 months from today when they have not even released the rest of FSD features yet.

This is your interpretation and you are injecting your own ideas and criteria to what he said. This is classic eisegesis. Elon already said in Q4 earnings call that NOA which was "FSD on the highway" was already "feature complete". Do you think NOA will wait on city driving before its performance and reliability improve? Of-course not, we have seen it by the fact that NOA has improved since its original introduction in Sept 2018.

Now if Elon had just said "hands on wheel detection for at-least another 6 months"

Then that to me would mean nothing. As I would believe Tesla was going the Supercruise route and will use the rear-mirror cam for driver check.

But its Elon's second statement that actually meant something. This is where he rejects driver monitoring, his conclusion was that the system is improving so fast, so much, that it will be completely moot, pointless and irrelevant and it will be irrelevant very very soon and very very quick.

Then he gave a timeline when driver supervision and intervention will not be needed. Infact he said that a driver intervening would actually DECREASE safety. So that's clearly a timeline where he believes that the system will be better than a human. Now what was that timeline?

"in fact, I think it will become very, very quickly, maybe and towards the end this year"


This corresponds to needing hands on detecting for atleast another 6 months.
towards end of this year = Q4 2019
hands detection for atleast another 6 months = Oct 2019

Someone saying that Elon portraying that he will have level 5 full autonomy in 6 months is absolutely right. To get technical you can replace 6 months to 6-9 months.