Tesla Autopilot scrutiny grows as U.S. asks if recall is planned

[Jan]

Tesla under recall scrutiny over autopilot issues

The U.S. National Highway Transportation Safety Administration stepped up its scrutiny of Tesla Inc.’s Autopilot, grilling the company over whether it should have issued a safety recall and questioning its rollout of new software customers are testing on city streets.

The regulator took issue with how Tesla deployed an update to its vehicles last month intended to improve their ability to detect emergency vehicles.

 

[Thor]

But God Elon can't do no wrong

 

[system32]

But God Elon can't do no wrong

Let's all chant
Lock him up...
Lock him up...
Lock him up...

From what I've read NHTSA has a new administrator that's anti-tesla which feeds in well with TSLAQ narrative and MSM.

Perhaps they should rename FSD to Driver Assist to get NHTSA off their back.

 

[jbrunevader]

Let's all chant
Lock him up...
Lock him up...
Lock him up...

From what I've read NHTSA has a new administrator that's anti-tesla which feeds in well with TSLAQ narrative and MSM.

Perhaps they should rename FSD to Driver Assist to get NHTSA off their back.

Won't matter; he implied that the Alzheimer patient in the White House was sleeping... Democrats if anything never forgive anyone who bursts their propaganda bubble.

 

[nivek]

Great news for my lidar stock!

 

[jbrunevader]

Great news for my lidar stock!

Lidar is no magic pill... it's highly prone to "false positives"; data that is impossible to interpret, hence Tesla has avoided it.

Simple reality is that it's easy to make a vehicle that never collides with anything... but nobody would buy it because it will never get out of the parking space.

The AI like a human needs to take actions based on what is considered a viable / recoverable risk -- where it has the potential to be better is that these type of decisions are made at micro-intervals far exceeding what any human could muster; far greater than any top formula 1 driver. Their biggest problem has been object recognition... but recent advances have made huge strides into overcoming even that.

It's no longer a case of if, but rather when they'll be far superior to any human driver in the same situation.

 

[Johnatan56]

Lidar is no magic pill... it's highly prone to "false positives"; data that is impossible to interpret, hence Tesla has avoided it.

Yes/no, there is a lot of research in that area, and it will probably be a combination of the two that wins out in the end. LiDAR's advantage of being useful through objects cannot be ignored imho. Also, the "false positive" rate depends on distance, within 20-25m it's usually quite good, e.g. https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9180326 which is testing a combination of LiDAR and camera, the combination they got an <1% false positive rate which is fine since you'd sort it out over multiple frames (you have lots of different data points). Note they're testing against snow, one of the things that LIDAR is the worst for, since it's solid but doesn't matter if car hits it, that's the <1% rate they got for detecting it.

 

[jbrunevader]

Yes/no, there is a lot of research in that area, and it will probably be a combination of the two that wins out in the end. LiDAR's advantage of being useful through objects cannot be ignored imho. Also, the "false positive" rate depends on distance, within 20-25m it's usually quite good, e.g. https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9180326 which is testing a combination of LiDAR and camera, the combination they got an <1% false positive rate which is fine since you'd sort it out over multiple frames (you have lots of different data points). Note they're testing against snow, one of the things that LIDAR is the worst for, since it's solid but doesn't matter if car hits it, that's the <1% rate they got for detecting it.

Who else but Tesla is leading in this area?

Do you really think Tesla would avoid lidar to just be difficult and / or avoid it for some petty reason.

There is a big practical reason why they avoid it; because it's impossible if not extremely difficult to use in the close quarter driving scenarios. If you've owned a car with some type of lidar then you'd know its completely too over sensitive and that strict adherence to its collision detection would render most cars immovable.

 

[wingnut771]

Would autopilot work at top speed on the autobahn?

/goes to check youtube

 

[Johnatan56]

Who else but Tesla is leading in this area?

Do you really think Tesla would avoid lidar to just be difficult and / or avoid it for some petty reason.

There is a big practical reason why they avoid it; because it's impossible if not extremely difficult to use in the close quarter driving scenarios. If you've owned a car with some type of lidar then you'd know its completely too over sensitive and that strict adherence to its collision detection would render most cars immovable.

Not sure what you're on about, there are a lot of cars with LiDAR, and I said it will be a combination of LiDAR and vision that will probably ultimately prevail. The main reason Tesla does not want to use it anymore is cost imho.
LiDAR works fine for e.g. parking lots, the issue is more while driving faster than that, it takes a lot of compute power to handle identifying objects in conjunction with vision.

 

[jbrunevader]

Not sure what you're on about, there are a lot of cars with LiDAR, and I said it will be a combination of LiDAR and vision that will probably ultimately prevail. The main reason Tesla does not want to use it anymore is cost imho.
LiDAR works fine for e.g. parking lots, the issue is more while driving faster than that, it takes a lot of compute power to handle identifying objects in conjunction with vision.

Nope... lidar has never been a matter of cost.

For anyone to look at a Tesla and say cost is why they avoid lidar is just laughable. Nope lidar is avoided because it's as granular as a PIR; the type of granularity that confuses AI; makes it unable to exit parking lots.

 

[Johnatan56]

Nope... lidar has never been a matter of cost.

For anyone to look at a Tesla and say cost is why they avoid lidar is just laughable. Nope lidar is avoided because it's as granular as a PIR; the type of granularity that confuses AI; makes it unable to exit parking lots.

But Musk has long insisted that he can deliver a self-driving car using cameras alone. At Tesla's Autonomy Day event in 2019, he called the technology a "fool's errand" and said "anyone relying on lidar is doomed." He also slammed lidar as "expensive sensors that are unnecessary."

Tesla is reportedly testing lidar sensors, which Elon Musk dismissed as 'expensive' and 'unnecessary' | Business Insider

Elon Musk says he can deliver on his self-driving promises without lidar. Tesla is reportedly testing the technology, sources told Bloomberg.

www.businessinsider.com

That is the main reason, now that they are cheaper Tesla is starting tests up again, that's my last comment on the matter.
You're talking about tens of thousands of dollars, even for a seventy thousand dollar car, that's not cheap.

Only in the last two/three years has it started coming down to hundreds of dollars instead, so it will end up being used.