Autonomous Car Progress

[diplomat33]

Ah, we agree. Tesla IS L2 (beta FSD included). Waymo is L4 in narrowly defined geofenced areas, yet still requires someone behind the wheel? And again, let's see how it does without safety drivers in those urban environments we talked about.

By definition, L4 means it does not need anyone behind the wheel in a specific, limited, ODD. That is why Waymo can remove the safety drivers in Chandler. The fact is that Waymo could remove all safety drivers now if they wanted to, it's just a question of how safe it would be.

And would you want some minimum wage earner being a "safety driver?" Next, you'll say the safety drivers are paid more than that. Which further weakens the economic viability of Waymo.

Not sure what the wage of safety drivers have to do with anything.

I disagree. Tesla is prototype L5 despite what they tell the CA DMV.

I think it makes sense to go with what the manufacturer says.

 

[Daniel in SD]

Nope. Not even close.

Obviously they're not close to meeting the safety level required but it's clearly their goal (which is what defines the levels according to the SAE).
How is FSD beta different from a test vehicle from any other AV company?

 

[Knightshade]

what production cars do tesla have that even come close to asil-d rating?

Again nothing in the SAE levels requires any particular asil-d rating

Not the first time this has been pointed out to you.

 

[diplomat33]

How is FSD beta different from a test vehicle from any other AV company?

One difference might the fallback. FSD Beta needs a driver to perform the fallback. AVs like Waymo do not need a driver to handle the fallback.

 

[rxlawdude]

One difference might the fallback. FSD Beta needs a driver to perform the fallback. AVs like Waymo do not need a driver to handle the fallback.

See SAFETY DRIVER. You forget this. Unless you're talking about the limited, several square mile suburban test where they have let them go without a safety driver. Very unlikely to happen in busy urban environments where demand for taxis is highest. And, Waymo does need a human to handle the fallback. It's that command center that can reprogram the path. Unfortunately, if you're near cones, it could result in significant delays till the in-person "backup" personnel arrive.

Comparing current functionalities is really still a distinction without a difference.

 

[mark95476]

I saw this post in the investor forum.... Just another "professional" analyst/firm providing -50% returns on their investment advice. You would have done better keeping cash in your mattress.

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

That is interesting and can be explained by differences in water chemistry, crappy design in Austin, or greater output in Austin. Given the TX mindset I am guessing crappy design but it could very well either of the other two. Its pretty obvious Texas will have higher output than Giga...

teslamotorsclub.com

Little has changed. People have been losing many billions doubting Tesla for years....

Tesla bears suffer record short-sale loss in 2020 - S3 Partners

Short bets on Tesla Inc were the most unprofitable trade in 2020, as the electric automaker saw its shares surge over 740% en route to becoming the world's most valuable automaker.

www.reuters.com

In a note to clients, Ihor Dusaniwsky, managing director of predictive analytics at S3 Partners, said mark-to-market short losses for Tesla totaled over $40 billion, as it was “far and away the most unprofitable trade in 2020 and had the largest yearly loss we have seen historically.”

Now Tesla is at >$1B/quarter profit and on the verge of opening two new factories.

If you don't think there's really people being paid to badmouth Tesla let me introduce you to one Gordon Johnson

 

[Bladerskb]

leaked video from a Xpeng City NGP beta tester.

https://twitter.com/x/status/1424936132547162116

 

[Bladerskb]

It’s funny how Tesla fans keep piling on HD maps and even Mobileye’s REM maps yet
Tesla will use Zenrin map in Japan which was entirely developed using Mobileye’s REM HD map technology. Tesla comes full circle at the feet of mobileye yet again. Lol

ofcourse the fans will excuse away the fact they keep relying on Mobileye’s tech

https://twitter.com/x/status/1424822434239336460

 

[Matias]

It’s funny how Tesla fans keep piling on HD maps and even Mobileye’s REM maps yet
Tesla will use Zenrin map in Japan which was entirely developed using Mobileye’s REM HD map technology. Tesla comes full circle at the feet of mobileye yet again. Lol

ofcourse the fans will excuse away the fact they keep relying on Mobileye’s tech

https://twitter.com/x/status/1424822434239336460

I don’t think that even the most hc Tesla fans would claim that Tesla manufactures everything themselves.

 

[Knightshade]

I don’t think that even the most hc Tesla fans would claim that Tesla manufactures everything themselves.

Yeah- Tesla has sourced maps from a number of places.

They're not generally the LIDAR-generated mm level ones Waymo relies on though.

That said- Japan as green notes in a later tweet is kind of weird in that they have some VERY VERY narrow, but technically legally public, streets.

Like streets a Model X physically can not fit driving down normally- but without maps that tell you street width data you can't know that in advance.

So by having that info in the maps, the car won't try routing you in a way that would put you in such a street.... (rather than simply relying on vision to notice the street is too narrow once you've already gotten there and it's potentially too late because you're there, with cars behind you)

 

[diplomat33]

Cruise put out a little marketing video of their car driving around SF with no safety driver:

AFAIK, Cruise is testing with no safety drivers now that they got a permit to do so.

 

[Knightshade]

I notice their self driving video contains surprisingly little self driving.

 

[Daniel in SD]

I notice their self driving video contains surprisingly little self driving.

Because watching self-driving is actually really boring when it's not failing and they're definitely not going to show that. Tesla is way better at viral content.
I don't really see the point of making marketing videos if you're not actually selling anything. Hopefully this means they're actually going to deploy soon...

+1 for The White Stripes though!

 

[powertoold]

I disagree. Tesla is prototype L5 despite what they tell the CA DMV.

This is interesting. I sort of agree with you. It's an aspirational level 5 feature in testing at level 2.

To my knowledge, it's also the only aspirational level 5 feature in development.

 

[Dan D.]

Because watching self-driving is actually really boring when it's not failing and they're definitely not going to show that. Tesla is way better at viral content.
I don't really see the point of making marketing videos if you're not actually selling anything. Hopefully this means they're actually going to deploy soon...

+1 for The White Stripes though!

They mentioned grocery delivery from Walmart in Scottsdale AZ in Nov 2020. Perhaps they'll offer that in San Francisco too.


I like how it's giving the pedestrian the "who wants to play chicken!" look

 

[diplomat33]

Motional is expanding their AV testing in CA:

The Motional team is excited to announce that we’re growing our West Coast operations. The expansion follows 18 months of high growth for Motional during which we closed the formation of our joint venture, extended our Asia operations to Seoul, and grew our headcount by almost 50%.

Motional made a deal with Lyft last year to launch driverless service in 2023:

This next move comes as Motional accelerates towards mass commercialization, which includes a landmark agreement with Lyft to launch a multimarket fully driverless service starting in 2023, and further establishes Motional as one of the world’s key AV players. We’ve had a presence in Los Angeles for more than five years, and are now increasing our investment in California with a new operations facility and road testing in the Los Angeles area, a new San Francisco Bay Area office, and extensive hiring that will more than double the size of our California team.

Motional Expands West Coast Presence | Motional

Motional harnesses deep industry experience in our mission to develop and deploy autonomous vehicles and to make driverless technology safe and reliable.

motional.com

I agree the "robotaxi race" is going to be interesting to watch in the next few years:

https://twitter.com/x/status/1425154943791493120

 

[rxlawdude]

Cruise put out a little marketing video of their car driving around SF with no safety driver:

AFAIK, Cruise is testing with no safety drivers now that they got a permit to do so.

Cruise was approved to test fully driverless cars (also called Level 4 in industry parlance) in California on October 15th. According to the DMV, Cruise can only test five driverless vehicles “on specified streets within San Francisco.” The vehicles are not allowed to exceed 30 mph, and can’t operate during heavy fog or heavy rain.

Yep, that's a real test. If Tesla operated at 30mph or less and not in rain, I bet it would work just as well if not better.

But 30mph? LOL!!!!

 

[Daniel in SD]

Yep, that's a real test. If Tesla operated at 30mph or less and not in rain, I bet it would work just as well if not better.

Tesla is by far the most capable autonomous vehicle, hypothetically speaking. I'm willing to bet that it will be at least a decade before the rest of the industry catches up.

 

[dgatwood]

This is an interesting issue. The two systems are independently analyzing the situation and therefore probabilities of failure are assumed to be random in each and therefore statistically uncorrelated. If that were clearly true the combined probability calculation would be defensible.

Even if you assume that, it's still indefensible. You touched on why in your last paragraph, but the short explanation is that not all failures are created equal:

• A once-per 10k mile failure resulting in it failing to see something will occur about once every 100M miles.
• A once-per 10k mile failure resulting in it seeing something that doesn't exist will occur about once every 5k miles.

And since those two cases are likely to be indistiguishable, you have to either trust the worst-case scenario and get phantom braking every 5k miles or trust the best-case scenario and get an accident every 10k miles. So the best-case overall system reliability is still bounded by the single sensor's 10k mile performance at best, or the 5k mile MTBF at worst.

 

[powertoold]

So the best-case overall system reliability is still bounded by the single sensor's 10k mile performance at best, or the 5k mile MTBF at worst.

The faster Tesla progresses, the faster the other fsd developers' houses of cards start to crumble. It'll be a sad day when we all realize they were just mostly hollow marketing companies; no offense to them though, as FSD is such a difficult problem, and they needed to keep the hype going to fund their development.

Of course, I can be wrong about all of this, but I'll definitely admit my mistake if so