Tesla

[NomisR]

As would I, but such cost cuttings measures, in a system with so much to prove, seems counterproductive.

https://www.forbes.com/sites/alanohnsman/2016/09/16/its-on-tesla-mobileye-spar-bitterly-over-what-triggered-autopilot-related-split/#3911da3a5ab0

Ammon Shashua told Reuters in an interview that his company was concerned Tesla was "pushing the envelope in terms of safety" in how it designed Autopilot.


Today In: Tech
Tesla responded to Shashua quickly, suggesting Mobileye tried to stop it from developing its own automated driving technology. "Tesla has been developing its own vision capability in-house for some time with the goal of accelerating performance improvements," the company said in an emailed statement. "After learning that Tesla would be deploying this product, Mobileye attempted to force Tesla to discontinue this development, pay them more, and use their products in future hardware."

Mobileye fired back Friday, accusing Tesla of lying. "Allegations recently attributed to a spokesperson for Tesla regarding Mobileye's position in respect of Tesla internal computer visions efforts are incorrect and can be refuted by the facts," Mobileye said in a statement. Regarding the assertion that "Mobileye was threatened by Tesla's internal computer vision efforts, the company has little knowledge of these efforts other than an awareness that Tesla had put together a small team," it said.

https://www.cnet.com/roadshow/news/mobileye-ceo-tesla-self-driving-cars/

"So what is Tesla stating? They are willing to take more risks in sensing mistakes because they have less redundancy, because they want to have a lower-cost technology." This, Shashua said, contrasts with a traditional OEM. "Another car manufacturer can come in and say, 'I want to have almost zero risk in having a sensing mistake and I'm wiling to pay more for that, or my customers are willing to pay more for that, so therefore I'm putting not only radars and cameras, but also lidars.'"

So yes, cost is part of the equation... but it seems more like Tesla playing fast and loose with what they can get away with as you have stated before than simply cost alone.

[r0tor]

Sounds like a BS spat about not installing (unneeded) lidar units that mobileye produces

[NomisR]

Sounds like a BS spat about not installing (unneeded) lidar units that mobileye produces

https://www.teslarati.com/tesla-self-driving-race-navigant-results-2019/

https://www.sae.org/news/2019/03/2019-navigant-autonomous-leaderboard

In case you bash the source, Teslarati is a bunch of Tesla fan bois..

Here's an article from 2018

https://www.greentechmedia.com/articles/read/report-ranks-tesla-last-for-automated-driving

It basically stated that Tesla took a step backwards... both Teslarati and greentechmedia are all pretty pro-Tesla.

[r0tor]

That's basically a ranking of vaporware

[NomisR]

That's basically a ranking of vaporware

And Telsa is ranking last or second to last in vaporware, and they're the ones that's selling it.

[r0tor]

And Telsa is ranking last or second to last in vaporware, and they're the ones that's selling it.

They are comparing Tesla systems which are driving in the real world to vaporware that may never get past driving on a closed course piloted by an engineer.

You can make the systems that are double, triple, quadruple redundant with a whole host of sensor and they would score well on that report... and also be so complicated and expensive to never actually get close to making it on a production vehicle.


The only way of making autonomous driving a reality is either remove human piloted cars completely from the road or use advanced AI to learn and predict the complete random acts of humans. It's something conventional coding and as many sensors as you want can never solve. When it comes to AI, the data set used to train the model is 10x more valuable then a bunch of extra sensors. The data set is going to have to come from real world miles - and Tesla is far in the lead at the moment.

[NomisR]

They are comparing Tesla systems which are driving in the real world to vaporware that may never get past driving on a closed course piloted by an engineer.

You can make the systems that are double, triple, quadruple redundant with a whole host of sensor and they would score well on that report... and also be so complicated and expensive to never actually get close to making it on a production vehicle.


The only way of making autonomous driving a reality is either remove human piloted cars completely from the road or use advanced AI to learn and predict the complete random acts of humans. It's something conventional coding and as many sensors as you want can never solve. When it comes to AI, the data set used to train the model is 10x more valuable then a bunch of extra sensors. The data set is going to have to come from real world miles - and Tesla is far in the lead at the moment.

GM Supercruise is Tesla's most direct and best competitor, and is ranked higher by reviewers because it's doing so responsibly, which is why people are criticizing Tesla.  Tesla is basically using owners as it's guinea pigs to beta test for them.   Programming can make up for the lack of sensors but it also needs considerable processing power which these cars currently don't have to do full L5 autonomous driving.  As I have posted previous, you're lookiing at a drop of 15-20% range with an electric vehicle currently to achieve that.  You can even look at things like how Google implements the Pixel phone that it tries to overcome with AI, and has done an excellent job in doing, but at the end of the day, they still had to add the sensors in subsequent generations in order to keep up. 

[r0tor]

Yea... Super cruise seems very well sorted...

https://www.theverge.com/2019/4/28/18515802/cadillac-super-cruise-ct5-ny-auto-show-2019

You can't wear sunglasses or use it on any road that hasn't been lidar mapped by GM... and only comes on 1 or 2 cars


Oh, and it doesn't use an onboard lidar sensor

[NomisR]

Yea... Super cruise seems very well sorted...

https://www.theverge.com/2019/4/28/18515802/cadillac-super-cruise-ct5-ny-auto-show-2019

You can't wear sunglasses or use it on any road that hasn't been lidar mapped by GM... and only comes on 1 or 2 cars


Oh, and it doesn't use an onboard lidar sensor

https://arstechnica.com/cars/2018/02/the-cadillac-ct6-review-super-cruise-is-a-game-changer/

Ambient light isn't an issue, because the steering wheel rim contains a number of infrared LEDs to illuminate you even in the dead of night, and DMS had no problems "seeing" my eyes even when I was wearing polarized sunglasses. If the DMS can't see you because its line of sight is blocked by an object, it reacts the same way as it would if it could see you but you weren't paying attention.

I'm sure there's issues, but having it disable rather than facing liability because the car crashed is a better alternative for the manufacturer.  Also 1 or 2 car.. oh yeah, 3 cars with Tesla.. big difference..

And why are you so invested in the argument?  You don't even own one.

[r0tor]

So the "big advantage" of super cruise is some flaky sensors monitoring the face of a driver to make sure they are looking at the road in a car that is suppose to drive itself... If you have to pay that much attention, why not just drive yourself

[giant_mtb]

So the "big advantage" of super cruise is some flaky sensors monitoring the face of a driver to make sure they are looking at the road in a car that is suppose to drive itself... If you have to pay that much attention, why not just drive yourself

Because, this:



plus stuff like this:

https://electrek.co/2020/02/19/tesla-autopilot-tricked-accelerate-speed-limit-sign/

is how people kill themselves and/or others.

[FoMoJo]

Yea... Super cruise seems very well sorted...

https://www.theverge.com/2019/4/28/18515802/cadillac-super-cruise-ct5-ny-auto-show-2019

You can't wear sunglasses or use it on any road that hasn't been lidar mapped by GM... and only comes on 1 or 2 cars


Oh, and it doesn't use an onboard lidar sensor

It uses cameras, radar, and LIDAR mapping data, combined with a robust driver monitoring system, to take a lot of stress and headaches out of highway driving.

I think it would stress me out and give me a headache. 

[Soup DeVille]

So the "big advantage" of super cruise is some flaky sensors monitoring the face of a driver to make sure they are looking at the road in a car that is suppose to drive itself... If you have to pay that much attention, why not just drive yourself

Isn't that exactly what Tesla says is required?

https://www.tesla.com/support/autopilot

[r0tor]

Because, this:



plus stuff like this:

https://electrek.co/2020/02/19/tesla-autopilot-tricked-accelerate-speed-limit-sign/

is how people kill themselves and/or others.

More of a statement of how well it works

[r0tor]

Isn't that exactly what Tesla says is required?

https://www.tesla.com/support/autopilot

Are you trying to say autonomous driving systems shouldn't be judged on how well they drive themselves, but how well they force the driver to act like they are driving when they are not?

[Soup DeVille]

Are you trying to say autonomous driving systems shouldn't be judged on how well they drive themselves, but how well they force the driver to act like they are driving when they are not?

I'm saying GM has the exact same requirements for driver involvement that Tesla does.

If you feel a need to make up things I'm saying to feel like you're winning, you aren't.

[r0tor]

I'm saying GM has the exact same requirements for driver involvement that Tesla does.

If you feel a need to make up things I'm saying to feel like you're winning, you aren't.

But the "benefit" is active eye monitoring to ensure you don't keep your eyes off the road for more then a couple seconds?  Such a great benefit in an autonomous car...

[Soup DeVille]

But the "benefit" is active eye monitoring to ensure you don't keep your eyes off the road for more then a couple seconds?  Such a great benefit in an autonomous car...

Here's the thing: neither are actually autonomous.


The benefit is the same as in every semi-automated car; to prevent human failures.

[Submariner]

Here's the thing: neither are actually autonomous.


The benefit is the same as in every semi-automated car; to prevent human failures.

The real "benefit" is that the technology is being pushed in a direction where eventually these systems can be truly autonomous.  Right now this is just a giant rolling proof-of-concept.

[Soup DeVille]

The real "benefit" is that the technology is being pushed in a direction where eventually these systems can be truly autonomous.  Right now this is just a giant rolling proof-of-concept.

I think both systems are making progress towards that goal, but neither can be thought of as autonomous yet, and thus need to ensure human input is available.

[Morris Minor]

Having witnessed a seven-car pileup on I-75 the other day - total carnage & chaos blocking the route for many hours, I'm happy for autonomous tech to step in to help.
Limited access freeways are the easiest and most obvious places to start this.

[AutobahnSHO]

...I'm happy for autonomous tech to step in to help.
Limited access freeways are the easiest and most obvious places to start this.

+1

[NomisR]

Definitely, I think there's a reasonable implementation also to mitigate liability and treating people as human guinea pigs and pushing the limits.. of course there's also the false advertisement in the naming scheme. 

[Morris Minor]

Maybe lanes dedicated to vehicles with a minimum of an adaptive cruise control & lane keeping system.
Musk Lanes.

[giant_mtb]

Maybe lanes dedicated to vehicles with a minimum of an adaptive cruise control & lane keeping system.
Musk Lanes.

Because roads aren't already clogged and under-constructed enough.

[MX793]

https://www.marketwatch.com/amp/story/guid/6AA12BB8-5803-11EA-BF64-5AC3A6D77C0D

[NomisR]

I was just about to post this.

[Morris Minor]

Because roads aren't already clogged and under-constructed enough.

You're sitting in lanes clogged with cars that are stationary because they're driven by humans. You stare enviously over at the Musk Lanes, where cars are moving smoothly & efficiently.

The problem would eventually take care of itself, and freeways will become 100% reserved for automated vehicles only.

[AutobahnSHO]

You're sitting in lanes clogged with cars that are stationary because they're driven by humans. You stare enviously over at the Musk Lanes, where cars are moving smoothly & efficiently.

The problem would eventually take care of itself, and freeways will become 100% reserved for automated vehicles only.

+1

[NomisR]

You're sitting in lanes clogged with cars that are stationary because they're driven by humans. You stare enviously over at the Musk Lanes, where cars are moving smoothly & efficiently.

The problem would eventually take care of itself, and freeways will become 100% reserved for automated vehicles only.

Yup, when all the cars are in communication with each other, it would technically minimize errors and maximize efficiency and space use.  We would get home much safer and faster than we otherwise would... theoretically anyways.