Elon Musk: EAP solved, on track for FSD completion in 2019 (No one else is close!)

[Tam]

...senators...

Ronald Reagan's quote "government is not the solution,...government IS the problem..." is getting old!

Since 2016, Michigan law has allowed autonomous vehicles that observe all traffic laws to roam freely in its state without a human nor steering wheel nor foot pedals. This is not about testing because numerous states have allowed testings already.

GM will make an autonomous car without steering wheel or pedals by 2019

The law is not the problem. The problem is today's technology is unable to keep up with basic traffic laws such as please don't crash onto a stationary object while cruising at 90MPH!

 

[mongo]

Call your senators and complain, they are the ones holding back L3 in the states. According to Amon there will be 2 million L3 cars running on Mobileye Eye4 in 2019.

But my point is, shouldn't Tesla software be more than .01%? If other companies are gunning for 2019,2020 and 2021. Shouldn't Tesla have released software that should replicate what these other companies have in-house? And for those that say Tesla is ahead. Then Tesla should have software that frankly trumps them all right?

Yet we have .01%? Something must be amiss.

Think about it. Time is running out for elon. We are 12 months from 2020. Other automakers timeline is about to catch up to him.
When this happens, what will Elon do? he can't promise anymore. It would be too shameful.

Nissan will launch more than 10 self-driving vehicles by 2020

Here's a description of Nissan's L3 Highway system for Japan that was delayed from 2018 into 2019

0.01% is your number, unless you have data to back it up.
Software requires appropriate hardware, since no delivered cars have AP 3.0, it seems a little difficult to say what Tesla's internal development branch software can or cannot do.

 

[Bladerskb]

0.01% is your number, unless you have data to back it up.
Software requires appropriate hardware, since no delivered cars have AP 3.0, it seems a little difficult to say what Tesla's internal development branch software can or cannot do.

FSD has to be able to go about one million miles before disengagement. By the way humans drivers have about one fatal crash every 100 million vehicle miles traveled. Let gp

Second of all Musk been hyping up FSD before AP3. Not just the Cross country demo but FSD features arriving early 2017. So no, its not a wait till the new hardware revision. You are basically saying AP2 hardware (40x more powerful) is useless and only good for replicating the features of AP1 that was produced using mobileye's 5 year old chip. That sounds rather convenient.

 

[Tam]

...hyping up FSD before AP3...

When not talking about today's capability, it's easy for anyone to write a science fiction for the future.

Elon Musk gave a mars lecture then Boeing replied back:

Boeing CEO Says Boeing Will Beat SpaceX to Mars

That's the same with others who said that they will beat Tesla with a Tesla Killer Car (because it's in future, not today) and so it's the same with when boasting about Autonomous Vehicles. It's in future so it's easy to boast about something that is not real today.

In the meantime, it's nice to learn what science fiction is based on.

 

[mongo]

FSD has to be able to go about one million miles before disengagement. By the way humans drivers have about one fatal crash every 100 million vehicle miles traveled. Let gp

2017 stats were 1.16 deaths per 100 million miles or one per 86 million, not that it matters.

Second of all Musk been hyping up FSD before AP3. Not just the Cross country demo but FSD features arriving early 2017. So no, its not a wait till the new hardware revision. You are basically saying AP2 hardware (40x more powerful) is useless and only good for replicating the features of AP1 that was produced using mobileye's 5 year old chip. That sounds rather convenient.

Don't conflate EAP and FSD. EAP was last reported to only need 2.x HW.

It there a problem with saying the generalized/ non-brittle FSD solution Tesla is developing is too processor intensive for AP 2.x?

 

[caligula666]

Don't conflate EAP and FSD. EAP was last reported to only need 2.x HW.

It there a problem with saying the generalized/ non-brittle FSD solution Tesla is developing is too processor intensive for AP 2.x?

Well, the point is that the AP2 hardware was supposed to be able to reach FSD based on Elon's 2016 claims and that this was only waiting for approval from authorities. The reality is that we know this was pure BS since not even AP2.5 can get there. Elon simply had no clue back in 2016 about what it would REALLY take to deliver FSD, it was just an engineering guess. And the same 2016 guesswork in all likelihood applies to all "FSD" subsystems (cameras, LIDAR-or-not, etc.).

So now, sure, we can have endless debates about "yeah the hardware should be enough" or "nope, it won't be", the reality is that we won't know for sure until... we know for sure. And the proof will be a Tesla car showing real autonomy, not the fun toy we are enjoying today.

 

[Bladerskb]

2017 stats were 1.16 deaths per 100 million miles or one per 86 million, not that it matters.



Don't conflate EAP and FSD. EAP was last reported to only need 2.x HW.

It there a problem with saying the generalized/ non-brittle FSD solution Tesla is developing is too processor intensive for AP 2.x?

Elon said that AP2 processing chip was more than enough to provide Level 5 FSD and repeated it up until summer of 2018.
Basically at this point you are revising history.

 

[Kruegmeister]

I don't get this. All the automakers and other companies already told you their timeline for their systems.
Ranging from 2019-2021. So saying where can i buy so and so is illogical.

...

When other companies start missing their timeline by 12+ months then you can start heckling them. Elon is already off by 12 months already.

Timeline: The future of driverless cars, from Audi to Volvo

Won't comment on Audi, etc cause I have no experience outside of Rentals...

I would doubt GM could do it because my Wifes Cadillac & my 2014 Corvette have Voice Commands that are so friggin bad that it's a complete joke. My Tesla understands if I say Call Zach & there is only a Zach it is smart enough to figure it out.
Not my Vette, it barely can understand the Call part. If you work for GM & worked on that System you should quit Software cause you suck. Period.
If you can't do something Amazon, Google & Apple + many more have done, then how do you think they have "experience" in software to make Self Driving...
Now Super Cruise seems decent, but its limited to pre-vetted freeways.
I like using mine on regular roads cause I don't get a ton of freeway driving.

OK, rant on Voice command by GM done... ughhh

 

[AnxietyRanger]

Again ... please, could you tell me which brand / model should I buy now, today, on 2018 for that level of autonomy? Please note that I live on a small town 20 miles from Madrid (Spain/Europe). Half of my daily commute are on secondary roads, where we have just one lane on each direction and the paintings are not very clear (although Autopilot interprets and drives them as good as I do).

I am ready to jump from Tesla to the company that autonomously drive this last 5% I am still driving manually on my daily commute.

Please, no demos, videos, dreams nor future availability, for that we already have the 2016 Tesla FSD demo which at this point we all agree it's just bullshit. Any demo or unavailable technology just goes to the bullshit bag, with the Tesla Demo.

But if you want, we can compare demos to demos. In that case, the 2016 Tesla demo is as impressive as imaginary.

It is true Tesla is the current leader on consumer-available autonomous driving systems, especially when one considers software updateable progress as a part of the equation - i.e. being part of the journey. This is unlikely to change anytime soon as Tesla's approach to consumer-availability is unique.

That said, putting all "demos" into the imaginary/BS basket is not correct either. As true as Tesla's lead in consumer-availability is, it is also true that companies in this area have a vast range of in-house progress on the self-driving car - and not all that progress is equal. We can argue who is leading in-house and who is not, but I guarantee you the progress is not equal.

Hence not all will reach the consumer-market at the same time. This latter point matters too, as Tesla's lead in consumer-availability can also be dependent on others not releasing their systems yet (i.e. someone launching a Level 4 feature tomorrow would change everything). That lead may well be affected once those other companies do release their solutions which are at various stages of progress differing from Tesla's stage of progress. (Not that I'm expecting Level 4 tomorrow or anytime soon, just an illustrative point.)

I'm of course simplifying above. Already someone looking for a hands-free solution might prefer Super Cruise, but overall I agree Tesla provides the most aggressive consumer-availability schedule at this time. At least until someone comes out with a consumer-available Level 3 or above system (car/car company responsible) - I would be rather surprised if that first someone is Tesla, but difficult to see the future is. We shall see.

Tag @lunitiks @croman @verygreen @S4WRXTTCS.

p.s. Reconciling @wk057 's latest post with Elon's quote on EAP being solved is a bit hard - but similarly Tesla's in-house progress might of course also differ on this from their consumer release: Navigate on Autopilot is Useless (2018.42.3)

Would be interesting to compare to what AP1 does after @wk057 post here.

Interesting video @verygreen. But AP1 does not have offer EyeQ3 a fisheye input (EyeQ3 does support it), so it will never be able to see what it would need to see to handle tight corners.

 

[SDRick]

I would be rather surprised if that first someone is Tesla, but difficult to see the future is. We shall see.

"Prediction is very difficult, especially if it's about the future."

 

[fmonera]

I don't get this. All the automakers and other companies already told you their timeline for their systems.
Ranging from 2019-2021. So saying where can i buy so and so is illogical.

Ok, so my options for 100% automatic daily commute:

1) Abandon Spain and go to live to a specific country (US) and not in any of the other 193 countries in the world. Find a house -and a work- so my daily commute involves some specific geofenced roads. I could buy a Cadillac then.

2) Go to live to one of the few high quality mapped citties in the world (instead of the other almost 2 million existing citties) and make sure my daily commute happens inside that citties. I still couldn't buy anything but I could use a L4/L5 Waymo taxi (only on good weather so that Lidar works correctly).

3) Jaguar I-Pace. I love this car, but unfortunately their L2 is not good enough.

4) Wait for automaker promises (including Tesla). 2019-2021 for their releases. Hopefully not geofenced to (part of) US, or the sollution will be garbage for me. Still will have to see how it works on real markets and real cars. Pass through v1 versions and bugs, on cars that don't upgrade themselves. And then, 2 years after release, have a decent solution (again, hopefully not geofenced).

5) Keep my Tesla which drives me 95% of my daily commute (secondary roads on Spain) without geofenced limits (except for very beta features such as NOA).

Any solution that relies on high quality mapping and geofenced information is irrelevant for me (and I suppose for any person too that lives outside the geofenced areas).

 

[mongo]

Elon said that AP2 processing chip was more than enough to provide Level 5 FSD and repeated it up until summer of 2018.
Basically at this point you are revising history.

1. It's software no one has done before. Not surprising to run into a HW limit. That does mean one expected to hit a HW limit.

2. He said the AP2.5 (not AP2.0) had the processing power (40x) needed around October 2016 All Tesla Cars Being Produced Now Have Full Self-Driving Hardware

3. However, Elon has hedged on HW since the beginning, thus the easily swappable computer. The camera suite has stayed the same (other than adding blue). Radar did get enhanced with 2.5 also. August 2017

“However, we still expect to achieve full self-driving capability with safety more than twice as good as the average human driver without making any hardware changes to HW 2.0,” the spokesperson said. “If this does not turn out to be the case, which we think is highly unlikely, we will upgrade customers to the 2.5 computer at no cost.”

Tesla has been working on a backup plan in case its self-driving promises fail

So not seeing the history history revision here. Just normal development. Tesla was always intending to provide the needed HW to FSD purchasers. Even camera swaps are not terrible to do, if needed.

Edit 4: April 2017 called out 2 years till FSD. https://electrek.co/2017/04/29/elon-musk-tesla-plan-level-5-full-autonomous-driving/
Not sure he ever said level 5 himself though. Police officers says: "I need to to turn around, go the wrong way up the on ramp, and then take a right on Petunia St (which is really Petunia Rd)" vs level 4 FSD that doesn't take verbal instructions.

 

[cwerdna]

I would doubt GM could do it because my Wifes Cadillac & my 2014 Corvette have Voice Commands that are so friggin bad that it's a complete joke. My Tesla understands if I say Call Zach & there is only a Zach it is smart enough to figure it out.
Not my Vette, it barely can understand the Call part. If you work for GM & worked on that System you should quit Software cause you suck. Period.
If you can't do something Amazon, Google & Apple + many more have done, then how do you think they have "experience" in software to make Self Driving...
Now Super Cruise seems decent, but its limited to pre-vetted freeways.
I like using mine on regular roads cause I don't get a ton of freeway driving.

OK, rant on Voice command by GM done... ughhh

GM bought Cruise Automation in March 2016. Cruise Automation had nothing to do w/the voice recognition on your GM cars.

You have seen Waymo and GM still lead the pack in California’s new self-driving report cards, right? You've also see the below for Waymo, Tesla and GM Cruise, right?
Autonomous Vehicle Disengagement Reports 2015
Autonomous Vehicle Disengagement Reports 2016
Autonomous Vehicle Disengagement Reports 2017

Background info at Testing of Autonomous Vehicles.

 

[emmz0r]

Ok, so my options for 100% automatic daily commute:

1) Abandon Spain and go to live to a specific country (US) and not in any of the other 193 countries in the world. Find a house -and a work- so my daily commute involves some specific geofenced roads. I could buy a Cadillac then.

2) Go to live to one of the few high quality mapped citties in the world (instead of the other almost 2 million existing citties) and make sure my daily commute happens inside that citties. I still couldn't buy anything but I could use a L4/L5 Waymo taxi (only on good weather so that Lidar works correctly).

3) Jaguar I-Pace. I love this car, but unfortunately their L2 is not good enough.

4) Wait for automaker promises (including Tesla). 2019-2021 for their releases. Hopefully not geofenced to (part of) US, or the sollution will be garbage for me. Still will have to see how it works on real markets and real cars. Pass through v1 versions and bugs, on cars that don't upgrade themselves. And then, 2 years after release, have a decent solution (again, hopefully not geofenced).

5) Keep my Tesla which drives me 95% of my daily commute (secondary roads on Spain) without geofenced limits (except for very beta features such as NOA).

Any solution that relies on high quality mapping and geofenced information is irrelevant for me (and I suppose for any person too that lives outside the geofenced areas).

I hate geofencing with a passion. Also. I bought FSD, then I will have FSD delivered. Or else they shouldn't have not sold it in Europe.

 

[Kruegmeister]

GM bought Cruise Automation in March 2016. Cruise Automation had nothing to do w/the voice recognition on your GM cars.

Testing doesn't say how much of that is Highway.
Super Cruise is mainly highway, not sure about their tests tho...
Highway is easier but higher stakes if you mess up due to speed.

You have seen Waymo and GM still lead the pack in California’s new self-driving report cards, right? You've also see the below for Waymo, Tesla and GM Cruise, right?
Autonomous Vehicle Disengagement Reports 2015
Autonomous Vehicle Disengagement Reports 2016
Autonomous Vehicle Disengagement Reports 2017

Background info at Testing of Autonomous Vehicles.

Good thing they bought, their development of tech has not impressed. Hope they don't ruin the company tho.
I've seen many smaller firms have the talent leave when the Big Corporate mind set of, we'll do it this way cause we're "inset bloated big name comphere"

 

[Kruegmeister]

GM bought Cruise Automation in March 2016. Cruise Automation had nothing to do w/the voice recognition on your GM cars.

You have seen Waymo and GM still lead the pack in California’s new self-driving report cards, right? You've also see the below for Waymo, Tesla and GM Cruise, right?
Autonomous Vehicle Disengagement Reports 2015
Autonomous Vehicle Disengagement Reports 2016
Autonomous Vehicle Disengagement Reports 2017

Background info at Testing of Autonomous Vehicles.

Another note is those disconnects don't show how much of that is Highway???
Unless the reports which I didn't open show more Data (on Tablet right now)?
Super Cruise was built for Highway which is easier. I'd hope they're trying to expand that.

My Tesla will get disconnected all the time because of:
Speed limit wrong and I don't wanna be rude to those behind me.
Just want to get around a slow guy.
Traffic Light with no car in front of me.
Guy is following too close and afraid of getting rear ended cause my car can stop so fast.
All that because I'm allowed to use on Streets other than Freeways, thank you Tesla I am enjoying the OTA aspect of your Model.

Granted it's not clear if this is independent drivers from same company driving different cars. Are some doing mainly Highway testing???

But I like all the people trying to push this forward.
I like the Tech and I like the Competition.
Maybe a company like Tesla, Google getting into Cars will keep GM from delivering the turd that is my Voice Command Systems they have me in 2014 & the Caddy one in 2017 was no better. I want it to work like Google Assistant, Samsung you can take Bixby & shove it!

 

[RDoc]

If the problem were the hardware then Tesla could just load a car up with a rack of processors, more cameras, radars, lidars, Tarot Cards, whatever, install the FSD software they've written, and demo it to get some real world experience. That would instantly shut up their critics, make the stock jump, and eliminate the "confusion" around the FSD option.

Maybe they're doing that somewhere in secret, but I doubt it.

 

[u00mem9]

If the problem were the hardware then Tesla could just load a car up with a rack of processors, more cameras, radars, lidars, Tarot Cards, whatever, install the FSD software they've written, and demo it to get some real world experience. That would instantly shut up their critics, make the stock jump, and eliminate the "confusion" around the FSD option.

Maybe they're doing that somewhere in secret, but I doubt it.

Exactly. This is the part I can never understand. This is the proof of concept that would surely be complete before a company would 1) spec the hardware, 2) launch production, and 3) start selling.

Like some here, I applaud Tesla’s willingness to let us drive quasi test cars, and I think that’s mostly a result of them having a lot less to lose in comparison to MB or GM. But I cannot understand defenders who don’t recognize that the known facts here betray the fact that Tesla did not, and does not know how to solve this problem. It’s insanity to continue to produce APx hardware with ever increasing computational power when they still don’t have it working in a test car.

 

[diplomat33]

If the problem were the hardware then Tesla could just load a car up with a rack of processors, more cameras, radars, lidars, Tarot Cards, whatever, install the FSD software they've written, and demo it to get some real world experience. That would instantly shut up their critics, make the stock jump, and eliminate the "confusion" around the FSD option.

Maybe they're doing that somewhere in secret, but I doubt it.

I think they are. There was a report awhile back that Tesla is recruiting employees into a FSD beta group:
Tesla reportedly starts incentivized Full Self-Driving beta test program for employees

 

[emmz0r]

I think they are. There was a report awhile back that Tesla is recruiting employees into a FSD beta group:
Tesla reportedly starts incentivized Full Self-Driving beta test program for employees

Hope they're not all Model 3 drivers, so they can give feedback in the X / S UI while they're at it