Autonomous Car Progress

[Bladerskb]

Nothing?

These would likely be the guys literatively testing very specific situations, and testing pre-public-release updates.

Distributing them to more places reduces some of the oft reported cases where it works great in say the bay area because that's where all their paid testers are but not elsewhere that it sees no testing until wide release.

Note the specific asks of:
attention to detail, and ability to work in a fast-paced dynamic environment.


So they might direct these folks "Test this situation, in these specific conditions, 100 times in a row" or "Test this situation, in these 10 different sets of conditions, 10 times a row" or even "Test this situation X times, then push this developer version button to switch software and retest X times, then reset with version C X times, etc...."

That's literally why safety drivers exist. So you are telling me the whole billions of miles of data is a sham? That just like they sent engineers to test Chuck's left turn for months.
You actually have to send people to test in very specific and repeated way? ZOMG!

 

[Bladerskb]

The day Tesla decides to remove the driver. When accidents and deaths of children start piling up the defense force from Tesla fans would be the most shameful and sad thing ever.

 

[Knightshade]

That's literally why safety drivers exist.

You appear to have confused "safety driver" with "software tester"-- those are different jobs and involve different things.

So you are telling me the whole billions of miles of data is a sham?

No, I told you the opposite of that.

I can explain it for you, but I can't understand it for you.

That just like they sent engineers to test Chuck's left turn for months.
You actually have to send people to test in very specific and repeated way? ZOMG!

If you have a specific case you want to test very specifically- yes.

If you need the car to read a speed limit sign you can just use fleet data, with a campaign that says "send me everything you think is a speed limit sign" or even "send me video from X seconds before the maps say there's a speed limit sign" and then you use that for training and labeling data.

Billions of miles come in very useful there.

But if you want to try out 50 different sets of parameters in 1 specific situation over and over, you can't easily do that with the fleet. You don't want 50 people who do something 1 time under a ton of different conditions and locations. You want 1 person who does the same thing 50 times in exactly the same location and conditions.

Now, the great thing about having both is once you have what you think is the right answer, you've got your massive fleet to push that answer to and let your billions of miles validate it to see if you need further iterative testing or if you got it right. These things work hand in hand- and these job posts aren't a new idea they've done it forever, they're just expanding the program to more places.

SURELY as a self proclaimed expert on this stuff you understand the difference.

 

[Knightshade]

The day Tesla decides to remove the driver. When accidents and deaths of children start piling up the defense force from Tesla fans would be the most shameful and sad thing ever.

Oh dear... he's gone full O'Dowd.

Never go full O'Dowd.


Anyway, in contrast to your FUD video from a random youtuber about AEB "not performing as expected" we have actual test data from actual safety testers showing the opposite.

Tesla Model Y wins 2023 safety award with "near-perfect safety rating"

Tesla's all-electric SUV has won leading Car of the Year Safety Award after record posting highest ever score.

thedriven.io

In 2022 the Model Y scored Euro NCAP’s highest ever safety score of 364 out of 400

They specifically cite how good AEB is in a Tesla in the story.

 

[spacecoin]

Anyway, in contrast to your FUD video from a random youtuber about AEB "not performing as expected" we have actual test data from actual safety testers showing the opposite.

They specifically cite how good AEB is in a Tesla in the story.

VW Diesel cars had practically zero emissions in tests too. So yes, I'd trust this youtuber over EuroNCAP and Tesla tbh.

Did Euro NCAP Absolve Tesla in Possible Code Cheat? No, It Washed Its Hands

More than saying Tesla did nothing wrong with its codes, Euro NCAP contradicted its own rules and admitted it was not able to check what happened.

www.autoevolution.com

"GreenTheOnly discovered that they were “running non-production code.” More than that: “100% of observed crash/ADAS tested cars in EU/U.S. have one-off builds on specially provisioned computers.”

Regardless if Tesla is cheating or not, I can't say I have any confidence in their AEB based on what I've seen in the real world. They only submit a few builds through testing, and none of these end up in consumer cars. This is undisputed.

 

[Knightshade]

Did Euro NCAP Absolve Tesla in Possible Code Cheat? No, It Washed Its Hands

More than saying Tesla did nothing wrong with its codes, Euro NCAP contradicted its own rules and admitted it was not able to check what happened.

www.autoevolution.com

"GreenTheOnly discovered that they were “running non-production code.” More than that: “100% of observed crash/ADAS tested cars in EU/U.S. have one-off builds on specially provisioned computers.”

Regardless if Tesla is cheating or not, I can't say I have any confidence in their AEB based on what I've seen in the real world. They only submit a few builds through testing, and none of these end up in consumer cars. This is undisputed.

*sigh*

This one was ALSO debunked.

Tesla Cleared All Doubts About Alleged Cheating In Euro NCAP Safety Tests

Strong accusations forced Euro NCAP authorities to investigate Tesla for alleged fraud in the European crash tests series. However, absolutely no trace of cheating in any possible way was detected by the European inspectors.

www.torquenews.com

Regardless if Tesla is cheating or not, I can't say I have any confidence in their AEB based on what I've seen in the real world. They only submit a few builds through testing, and none of these end up in consumer cars. This is undisputed.

Undisputed?

If consumers never see the code how did Green see it?

 

[Bladerskb]

Oh dear... he's gone full O'Dowd.

Never go full O'Dowd.


Anyway, in contrast to your FUD video from a random youtuber about AEB "not performing as expected" we have actual test data from actual safety testers showing the opposite.

Tesla Model Y wins 2023 safety award with "near-perfect safety rating"

Tesla's all-electric SUV has won leading Car of the Year Safety Award after record posting highest ever score.

thedriven.io

They specifically cite how good AEB is in a Tesla in the story.

lol typically Tesla fan defense force logic, classify anything that is remotely negate to Tesla as "FUD"!!!! This is FSD BETA running into pedstrains...
IF this was any other system, you Tesla fans would preach it from the house top. Shows yet again for the 1 million times, you tesla fans don't care about evidence you want propaganda.

 

[Knightshade]

lol typically Tesla fan defense force logic, classify anything that is remotely negate to Tesla as "FUD"!!!! This is FSD BETA running into pedstrains...
IF this was any other system, you Tesla fans would preach it from the house top. Shows yet again for the 1 million times, you tesla fans don't care about evidence you want propaganda.

What's especially funny here is you just got done being mad that Tesla was previously using regular folks to "test" their ADAS systems and that pros are really important.

But when it's pointed out professional testing debunked your regular folk youtube video- showing under rigorous professional testing that safety systems (with AEB specifically mentioned) in a Tesla is superior to any other car- that's ALSO bad.

 

[zoomer0056]

Ah, but it's also the word of VW!

You know, the folks who in 2013 told us they'd be the leader in electric cars by 2018!

We already know who the EV leader is. According to scary Mary and Biden, GM is the EV leader.

 

[Bladerskb]

You appear to have confused "safety driver" with "software tester"-- those are different jobs and involve different things.

Safety drivers in other companies perform the responsibilities listed in Tesla's listing.

No, I told you the opposite of that.

I can explain it for you, but I can't understand it for you.

I care what the facts are. And the facts point to billions of miles of data being only useful for validation and mapping. Other than that its a complete sham. This has been proven from the result of the last 8 years of Tesla's Autonomous Driving development. Even Tesla agrees with this in their Talks. When you compare Tesla's product with other products the truth is also there. Billions of miles of data is not needed to create the same product with similar or superior quality.

If you have a specific case you want to test very specifically- yes.

If you need the car to read a speed limit sign you can just use fleet data, with a campaign that says "send me everything you think is a speed limit sign" or even "send me video from X seconds before the maps say there's a speed limit sign" and then you use that for training and labeling data.

Billions of miles come in very useful there.

This is useless and completely false. You don't need billions of miles to build a great perception system which is literally proven by other systems like Mobileye, Huawei, Xpeng, etc. Again this is why we need to compare PR statements with actual facts. We need to bring Tesla's PR statement into reality.
The billions of miles narrative is literally a lie.

But if you want to try out 50 different sets of parameters in 1 specific situation over and over, you can't easily do that with the fleet. You don't want 50 people who do something 1 time under a ton of different conditions and locations. You want 1 person who does the same thing 50 times in exactly the same location and conditions.

This is literally what safety drivers DO!

Now, the great thing about having both is once you have what you think is the right answer, you've got your massive fleet to push that answer to and let your billions of miles validate it to see if you need further iterative testing or if you got it right. These things work hand in hand- and these job posts aren't a new idea they've done it forever, they're just expanding the program to more places.

I have from day 1 (back in 2016) in the forum always said, Mapping and validation are the only benefits of immensely large fleets.

 

[diplomat33]

Here is a map of the US that shows which States allow AVs:

As we can see, most US States allow AV deployment. There are only a few exceptions.

 

[Knightshade]

I care what the facts are.

Do you though?

You're the guy who despite every safety and regulatory testing body in the world finding Teslas AEB and other safety systems the best in the world are posting a youtube video about how some rando got their car to hit a beer keg in the road.

I have from day 1 (back in 2016) in the forum always said, Mapping and validation are the only benefits of immensely large fleets.

From day 1?

You- literally today

So you are telling me the whole billions of miles of data is a sham?

I care what the facts are. And the facts point to billions of miles of data being only useful for validation and mapping.

I mean I guess this is progress given you suggested a few posts ago it wasn't useful at all.

Don't worry- you'll get there eventually!

 

[Bladerskb]

From day 1?

You- literally today

I mean I guess this is progress given you suggested a few posts ago it wasn't useful at all.

Don't worry- you'll get there eventually!

Oh typically Knightshade that takes people's statements out of context and tries to twist them.
This is why you couldn't respond to the actual facts. Because no one can.
But go ahead, keep parroting the narrative that Tesla's billions of miles of data puts them 10 years ahead.
I'm not gonna waste time with you on this.

 

[spacecoin]

This one was ALSO debunked.

There is nothing in that article that wasn't in the one I linked. Are you claiming Tesla submits every build throrugh external testing now?

If consumers never see the code how did Green see it?

He got hold of a computer from an NCAP test i think.

 

[Bladerskb]

VW with Mobileye Drive is starting testing in Austin, plans to grow test fleet to at least 4 more US cities in coming years. VWGoA plans to launch commercial AVs (powered by Mobileye Drive) in Austin by 2026. For those who might not know, Mobileye Drive is Mobileye's autonomous driving stack designed for "driverless" like robotaxis and driverless delivery.

https://twitter.com/x/status/1676952562782863362

This actually puts Mobileye around 6 years behind Waymo on the leaderboard.
For those who think anyone can do what Waymo is doing, all you have to do is geofence.
Even Cruise is roughly 2 years behind where Waymo is and Zoox is around 4 years behind Waymo.

Looks like Waymo is the one with the unsurmountable lead (as long as they don't screw up).

 

[diplomat33]

This actually puts Mobileye around 6 years behind Waymo on the leaderboard.
For those who think anyone can do what Waymo is doing, all you have to do is geofence.
Even Cruise is roughly 2 years behind where Waymo is and Zoox is around 4 years behind Waymo.

I agree with your assessment. My only caveat would be that I think Mobileye will focus more on consumer cars, not robotaxis like Waymo is doing. I say this because their main focus seems to be SuperVision and Chauffeur as we've seen in the recent OEM deals that they announced. And, Mobileye has made it clear that they are not interested in personally managing a driverless fleet, like Waymo is doing.

And I suspect Mobileye will focus more on advanced ADAS, ie "eyes on/hands-off" rather than "eyes off" or "driverless". It will be easier to deploy with carmakers since the liability will remain on the driver. "Eyes off" or "driverless" is much harder to deploy, especially with carmakers who will likely be hesitant about assuming liability. In a few years, Mobileye will probably deploy some limited "eyes off", ie highways. A limited ODD will make it easier to manage risk and liability. But for now, they will focus on "eyes on/hands off". And that's ok since they can generate lots of profits with ADAS. Mobileye is one of the rare AV companies that actually does not need to do driverless to make a profit.

Looks like Waymo is the one with the unsurmountable lead (as long as they don't screw up).

Agreed. I don't see anyone catching up to Waymo at this point. In fact, Waymo just added another 45 sq mi to their metro Phoenix area. They seem to be scaling regularly now.

https://twitter.com/x/status/1677316564238888963

And if you look at the AV regulation map I just posted, I think Waymo would be wise to focus outside of CA. There are lots of States with friendly AV regulations. I think Waymo could scale much faster in those States. If they put all their eggs in the CA basket, they are just going to get bogged down by these anti-AV groups that will delay and block their deployments. And Waymo is showing that they have a robust stack that can do safe driverless in a meaningful ODD. There are plenty of cities in those pro-AV States where Waymo could flourish.

 

[Knightshade]

Oh typically Knightshade that takes people's statements out of context and tries to twist them.
This is why you couldn't respond to the actual facts. Because no one can.
But go ahead, keep parroting the narrative that Tesla's billions of miles of data puts them 10 years ahead.
I'm not gonna waste time with you on this.

Man you write a ton without ever actually saying anything.... probably for the best given when you DO try to say something it's easily debunked with a link or two

Protip- if you don't want people quoting your own words being self-contradicting, try to maintain a more consistent set of FUD statements.

 

[stopcrazypp]

Man you write a ton without ever actually saying anything.... probably for the best given when you DO try to say something it's easily debunked with a link or two

Protip- if you don't want people quoting your own words being self-contradicting, try to maintain a more consistent set of FUD statements.

I find it incredibly funny him accusing people of taking his statement out of context and twisting it, when that is a tactic that he ironically commonly employs and many have been a victim of in this forum.

You beat me to it, but using testers does not invalidate the usefulness of billions of consumer miles (something he later admits), but yet he found it necessary to make fun of it. An analogy I thought of was supercharging vs home charging. Supercharging is incredibly useful for certain scenarios, but that doesn't make home charging any less useful, as it powers a bulk of the charging people do.

Ironically, it is a long running argument by Tesla naysayers that have insisted Tesla's releases are untested other than by end users and thus DANGEROUS, when Tesla fans know there is internal testing before release to the general public. Besides from the points you raised, there will always be demand for testing internal releases, given consumer testing will never be able to accomplish the same thing (they simply don't have access to all releases).

 

[stopcrazypp]

Looks like people in SF have found a new way to make their voices heard ahead of the CPUC meeting on AVs: crippling the fleet by putting traffic cones on their hood.

Fed Up Citizens Disable San Francisco's Self-Driving Cars With Traffic Cones

An activist group called Safe Street Rebel are promoting the tactic on TikTok.

jalopnik.com

 

[diplomat33]

Looks like people in SF have found a new way to make their voices heard ahead of the CPUC meeting on AVs: crippling the fleet by putting traffic cones on their hood.

Fed Up Citizens Disable San Francisco's Self-Driving Cars With Traffic Cones

An activist group called Safe Street Rebel are promoting the tactic on TikTok.

jalopnik.com

This is so dumb. And it will not make SF safer. On the contrary, it will only cause more stalls and cause more traffic issues. If they really cared about resolving stalls, they should work with Waymo, not try to pull stupid tiktok challenges.