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

[whitex]

AP2, NO. AP3 with the new chip, under non-severe weather conditions as you mentioned, YES, I do think, with the right software that it will be able to do a 200 mi FSD summon. Laugh all you want, but I think so, again, with the right software.

There are many complex problems Tesla is not addressing, but let's start with a very simple one. When the car drives, the cameras around get dirty (see what happens to the read view camera in rain). So a simple question for you, how is this right software, that you are suggesting can be written, going to keep the cameras clean? Or do you define non-severe weather as perfect sunshine, driven on perfectly clean roads with no water, dust, or anything that could possibly get kicked up and settle on the cameras?

 

[whitex]

He's running a business, he's trying to make money, if he didn't over promise and under deliver, Tesla wouldn't' be where they're at now.

Do you think Elon accepts others taking his money and under-delivering or not delivering at all? Do you think Elon would give me a brand new P100D if I give them a check capable of covering the price but limited by my account balance (akin to the car he sold me which is capable of 691hp but limited by the battery which would have to deliver an additional 50% power to reach 691hp, of course just like Tesla didn't fess up to it for 2 years, I wouldn't tell them for 2 years that my account is limiting the check)?

 

[timvracer]

Elon constantly says things that he can't live up to, oh well.
But the Tesla is still awesome, and I know Blader keeps claiming how other manufacturers are ahead of Tesla.. but tell me today what car I can go buy.. TODAY.. that can do what the Tesla is doing now.

crickets.

 

[tonglaji]

There's a difference between projecting ambitious goals that you think you can reach but don't due to unforeseen circumstances and projecting ambitious goals that you know are false in an effort to drive sales/whatever. I don't for one moment believe Elon is doing the second half of that. I genuinely believe he projects what he thinks but because he can't see every obstacle on that path, his projections routinely come up short. I'd rather have ambition than apathy...

That being said, NoA is a total and complete joke and as I've said in other threads, if this is indicative of Tesla's progress with FSD then oh boy... We're a looooonnnnnggggg ways away from anything resembling FSD... I've stopped using NoA all together because it's that bad. It could be made much better by simply disabling the stupid autosteer speed restrictions that the car gets handicapped with on freeway on/off ramps...

Jeff

There is no need for such negativity on a person who we agree is a visioner and a company we all like. And we all know too well to disregard the timelines Elon announces. There is currently no other car that can match the Tesla capabilities, and Tesla as a company is our best bet to get our hands on some FSD features in the near future. We should be proud that we are living through the infancies of FSD, and it will be a good story to tell 30 year later, if we are still around.

 

[whitex]

There is no need for such negativity on a person who we agree is a visioner and a company we all like. And we all know too well to disregard the timelines Elon announces. There is currently no other car that can match the Tesla capabilities, and Tesla as a company is our best bet to get our hands on some FSD features in the near future. We should be proud that we are living through the infancies of FSD, and it will be a good story to tell 30 year later, if we are still around.

The problem is that it's not just timelines. Do you think my P85D will ever magically get a 50% power boost to reach the advertised 691hp? Or do think AP1 will ever "find me anywhere on private property" like Elon promised?

Or do you consider never delivering simply a timeline issue where time=infinity? So according to you Elon, never breaks any promises, he's just infinitely late?

 

[diplomat33]

So a simple question for you, how is this right software, that you are suggesting can be written, going to keep the cameras clean?

Software won't clean the cameras but software can mitigate the problem. There is software that exists today that compensates for optical distortions caused by water and other particles. The software analyses each frame and can eliminate the distortions post-production. Astronomers use it to eliminate atmospheric distortions. So this software would allow the car to continue self-driving when the cameras are just mildly dirty. But you would also use software to analyse when your cameras get so dirty that they are not able to see well enough and notify the driver. In fact, V9 already has this. When driving in severe weather, AP will notifies the driver that severe weather will limit AP functionality. And a FSD car would be able to pull over on its own when it detects that it is not able to see well enough.

so software won't clean the cameras but it can address the problem.

PS: I trust Tesla engineers have thought of this problem. They are some of the best in the industry. So I highly doubt that they spent all these years developing EAP and FSD only to get stumped in the end by a simple question like what to do when the cameras get dirty.

 

[whitex]

So this software would allow the car to continue self-driving when the cameras are just mildly dirty.

I agree with you there, but a cross country summon is likely to get the cameras beyond mildly dirty. Not to mention sudden splash with mud scenarios where cameras are blinded instantly.

But you would also use software to analyse when your cameras get so dirty that they are not able to see well enough and notify the driver.

Wait, what driver? If I'm summoning the car from across the country as per Elon's promise, there are no humans in the car. You'd think in the cross country summon the cameras are going to be more than mildly dirty,

I trust Tesla engineers have thought of this problem. They are some of the best in the industry.

Lol, you mean how they thought through how to use parking sensors for BSM, instead of using proven radar sensors? Or maybe how the car was supposed to find me anywhere on private property without having anything but one camera and ultrasonic sensors which can't detect surface drops, or even a duffel-bag in front of the car. Or are you telling me some time before end of time there will be some software update that will make ultrasonic sensors work well at highways speeds with car air-speeds up to 90mph (Elon physics the sound will travel though ether instead of air?), and my AP1 car will grow some extra-sensory software based perception to sense things so it can safely find me anywhere on private property?

 

[The Duke]

Not only does every other company deploy NN on their lidar and camera. They are far more advanced and sophisticated than Tesla.

Other than Mobileye, happen to know what chipset these systems run?

 

[RDoc]

Software won't clean the cameras but software can mitigate the problem. There is software that exists today that compensates for optical distortions caused by water and other particles. The software analyses each frame and can eliminate the distortions post-production. Astronomers use it to eliminate atmospheric distortions. So this software would allow the car to continue self-driving when the cameras are just mildly dirty. But you would also use software to analyse when your cameras get so dirty that they are not able to see well enough and notify the driver. In fact, V9 already has this. When driving in severe weather, AP will notifies the driver that severe weather will limit AP functionality. And a FSD car would be able to pull over on its own when it detects that it is not able to see well enough.


so software won't clean the cameras but it can address the problem.

Um, not quite. The astronomical systems work by having an artificial star created by a laser that they can track and compensate pretty minor distortions with. They can't correct for a cloud for example. Having a drop of water or a bunch of dirt on the lens makes a permanent distortion and/or blackout that is pretty much impossible to correct because the system doesn't know what it should look like. The data just isn't there.

PS: I trust Tesla engineers have thought of this problem. They are some of the best in the industry. So I highly doubt that they spent all these years developing EAP and FSD only to get stumped in the end by a simple question like what to do when the cameras get dirty.

That is a VERY questionable statement. Google (Alphabet) and Mobileye would seem to be much more likely candidates for the best in the industry. Considering that Tesla has yet to conquer the heights of USB software, e.g. playlists, much less automatic braking for pedestrians etc. I'm not all that impressed by their software prowess.

 

[Mo City]

There are many complex problems Tesla is not addressing, but let's start with a very simple one. When the car drives, the cameras around get dirty (see what happens to the read view camera in rain). So a simple question for you, how is this right software, that you are suggesting can be written, going to keep the cameras clean? Or do you define non-severe weather as perfect sunshine, driven on perfectly clean roads with no water, dust, or anything that could possibly get kicked up and settle on the cameras?

This is the fundamental question nobody has answered yet. Radars on all 4 corners (or at least a rear radar) are required for legit FSD.

 

[diplomat33]

Um, not quite. The astronomical systems work by having an artificial star created by a laser that they can track and compensate pretty minor distortions with. They can't correct for a cloud for example. Having a drop of water or a bunch of dirt on the lens makes a permanent distortion and/or blackout that is pretty much impossible to correct because the system doesn't know what it should look like. The data just isn't there.

Well dirt, I agree. Dirt is solid so yes it would permanently block the image. But a drop of water refracts and reflects light which obeys known laws. You can compare the distorted image to the image you know should be getting to see how the water is distorting the image.

 

[djgarrett21]

Well dirt, I agree. Dirt is solid so yes it would permanently block the image. But a drop of water refracts and reflects light which obeys known laws. You can compare the distorted image to the image you know should be getting to see how the water is distorting the image.

I'm glad some hold engineers and their abilities in such high regard.

Dirty cameras isn't a use case for the FSD hardware currently installed on Tesla vehicles.

Realistically FSD in inclement weather is going to require the Enhanced FSD and HW5 (or maybe HW6)

 

[Twiglett]

This is his way of rubbishing everything that Tesla is doing while attempting to show that his employer (mobile eye) is really truly just about to release proper FSD honestly.

 

[RDoc]

Well dirt, I agree. Dirt is solid so yes it would permanently block the image. But a drop of water refracts and reflects light which obeys known laws. You can compare the distorted image to the image you know should be getting to see how the water is distorting the image.

Right, but that's exactly the problem. The car doesn't know what the image should be. If it did, it wouldn't need the camera. We're talking about a very dynamic scene and natural rain, not a test pattern and some carefully applied ISO standard water drop.

To really do cross country autonomous FSD, i.e. the oft promised summon from LA to NYC, the car has to be able to keep the sensors working. For a human, if it's raining, or slush, dirt, etc. gets on the windows, they can run the wipers and/or washers. Teslas have no equivalent for the cameras or other sensors.

And we haven't even begun to talk about sun flares, fog, heavy snow, etc.

 

[RDoc]

This is his way of rubbishing everything that Tesla is doing while attempting to show that his employer (mobile eye) is really truly just about to release proper FSD honestly.

Elon works for Mobil Eye? Do you have anything to back that statement up?

 

[buttershrimp]

Blader has been reactivated by Roger Stone!! It's amazing that with each step towards FSD, you are still trying to convince yourself it isn't going to happen.

Comon' dude. It's happening.

 

[mongo]

Right, but that's exactly the problem. The car doesn't know what the image should be. If it did, it wouldn't need the camera. We're talking about a very dynamic scene and natural rain, not a test pattern and some carefully applied ISO standard water drop.

To really do cross country autonomous FSD, i.e. the oft promised summon from LA to NYC, the car has to be able to keep the sensors working. For a human, if it's raining, or slush, dirt, etc. gets on the windows, they can run the wipers and/or washers. Teslas have no equivalent for the cameras or other sensors.

And we haven't even begun to talk about sun flares, fog, heavy snow, etc.

The 3 front cameras have their own defroster and are within the area swept by the wipers.

 

[Bladerskb]

Sorry @Bladerskb but I think you are going against the current. Tesla is making progress towards FSD each time they release a new update. Nav on AP may have issues but it works far better than 5%. And each time, you insist that Nav on AP doesn't work, Tesla will release an update that makes it just a little bit better. Yes, Tesla still has a lot of work to do to reach FSD but they are making progress. And, you are right to be skeptical about timelines. Yes, it is very possible that Musk is wrong about timelines. But the bottom line is that Tesla is moving in the right direction towards FSD and they are making progress. So insisting that they are not making progress, is just silly. You claim "Tesla can't do X" and then Tesla does X. You insist "But Tesla won't do Y" and then Tesla will do Y. "But, but, Tesla can't do Y good enough" and then Tesla does Y better and better. You are just fighting against the inevitable. Eventually, Tesla will get to FSD.

And do you work for MobilEye or something?? I am just curious because it seems like each time Tesla releases an update that makes EAP a little bit better, you basically post that "Tesla sucks and MobilEye is better".

Possible? you just read about him missing 5 timelines LOL.

I'm not fighting the inevitable. I have laid out exactly what Tesla can do and when it will be able to achieve it and i have been Bulls-eye so far. Think about it. NOA was Tesla's general approach to highway autonomy yet it works 5% of the time. Infact as i outlined the only new thing it does is change lanes to take exits when next to an exit lane, slow down to a complete stop to change lane (which is dangerous).

The success factor is the problem. It only changes lanes once per mile, it constantly misses exits. It sometimes takes the wrong fork, sometimes dive for the divider. It 100% will crash during interchange merge with another car. I mean come on. does that sound like general highway autonomy to you?

Highway autonomy is EASY. especially if you want to release it as a L2 ADAS system. Delphi completed a cross country drive in 2015. that's almost 4 YEARS AGO! The fact Tesla is still struggling and NOA is their best effort is alarming!

Random startups spun every other day and can do highway autonomy with their eyes closed.

Here's what i said about EAP NOA initial release 2 years ago.

And about automatic lane change, it will be disappointing because it won't any different than whats happening today. Infact AP1 car could in theory do it. It won't change lanes in low speed/traffic jams and it won't be doing processing such as... is there obstruction/stopped car/barrier/lane closure upcoming in the lane that i want to turn into? Basically it wont monitor the environment and will be prone to too many mistakes to be useful.

Also as you mentioned freeway transition. when merging into a new freeway and also when changing lane while on the ramp and having to interact with other cars in order to get to the lane to merge into the new freeway. all these things wont be purpose.
its literally a glorified toy that you have to monitor moment by moment.

My point is, if NOA is Tesla's general highway autonomy, what do you think Elon's so called level 5 general full self driving is gonna look like next year? Elon is still setting up lemonade stands laced with kool-aid.

 

[Bladerskb]

I agree with you there, but a cross country summon is likely to get the cameras beyond mildly dirty. Not to mention sudden splash with mud scenarios where cameras are blinded instantly.


Wait, what driver? If I'm summoning the car from across the country as per Elon's promise, there are no humans in the car. You'd think in the cross country summon the cameras are going to be more than mildly dirty,

Not just that but simply everyday driving will cause this. Imagine driving through muddy water after a heavy rain and the cars next to you sprays your car and all of a sudden both your repeaters and rear-view cams are disabled.
Thanks for being one of the few thinking about this logically.

 

[S4WRXTTCS]

There goes Blader again thinking most of us believe Elon.

It's not that most of us believe Elon, but we understand this is a journey.

To be honest this whole autonomous driving thing always feels just a couple years away. It doesn't matter if it's Tesla, Waymo, Cruise, etc,.

It's beyond hyped by everyone.

Blader himself posted a video of a Waymo being unable to merge, and having to stop at the end of the merge point.

It's not just the companies involved in autonomous cars, but regulation is a complete mess as well. Then we have a serious infrastructure problem where even humans have difficulties in some situations,

We really have to scale back our expectations.

The difference between Tesla and everyone else is you can at least experience part of the journey with a Tesla versus waiting on the side lines watching tech demo after tech demo.