Autonomous Car Progress

[Jeff N]

Cruise is now 24/7 in all of SF.

A later tweet in that thread shows this image:

Here’s what that really means as I understand it. They have apparently tested the cars with human supervision across the full city map shown in Kyle’s first tweet and found the current system to be sufficiently ready.

For actual “driverless” use, they are currently testing the “Cruisers 24/7” map shown in the later tweet with a group of employee test riders. The “Power Users 24/7” daytime and nighttime maps show the areas currently available to a group of non-employee test riders.

I’m a member of the Power Users group and have been taking some daytime rides since they became available to me a couple of weeks ago.

I plan to post something here about my daytime experiences with Cruise in the next day or two.

 

[stopcrazypp]

In context, I meant it is meaningless for removing driver supervision (ie L4 autonomy). It is obviously not meaningless as a stat for the safety of a L2 driver assist system.

Having said that I still think the stat is very misleading. Tesla only measures accidents with airbag deployment or other active restraint employed. So Tesla is only measuring serious crashes. But they measure that accident rate against the overall industry average which includes all police reported accidents and vehicles of different years, some with ADAS and some without. So it is not an apples to apples comparison. Additionally, city driving is at lower speeds so we would expect less of a chance of a crash involving airbag or active restraint deployment, especially if the driver is attentive and taking over before FSD gets into trouble or not using FSD Beta in risky situations in the first place. So the stat is very stacked IMO.

Really, the only thing that the stat says is that an advanced driver assist with an attentive driver will have fewer serious crashes than the average of all accidents. Well, duh, that is not surprising.

What you talk about only affects the NHTSA stat, it doesn't affect the other Tesla vehicle stat which is also posted. As for your last claim, actually there are plenty of naysayers that argue otherwise. That AP is more dangerous because the driver when using it might not be attentive (Tesla isn't sampling only the attentive ones).

 

[diplomat33]

That AP is more dangerous because the driver when using it might not be attentive.

I think that was always an overly simplistic argument. Certainly, if the driver fails to pay attention at the key moment when the automated driving system is about to crash then yes, it will be more dangerous. Sadly, Tesla has had some fatal crashes where the driver was not paying attention at the wrong time. But the driver could not pay attention at a time when the automated driving system is perfectly fine and that won't make the system more dangerous. And you could have a very unreliable automated driving system that forces the driver to pay more attention because the system is so unreliable and that would not be a more dangerous system either. Lastly, you could get to the point where the automated driving system is so reliable that it does not need driver supervision anymore and then it won't matter if the driver is not attentive. That would not be a more dangerous system just because the driver is not attentive.

 

[MP3Mike]

To hit Elon's stated goal of 10x safer than a human, they need to get down to 0.15 accidents per 1m miles without supervision. FSD beta drivers are supposed to supervise at all times.

But that goal doesn't apply to HW3. The goal Elon stated for HW3 is 2-3x safer than a human.

 

[EVNow]

I think that was always an overly simplistic argument.

Right.

The way to think about it is
- Driver fault (including not paying attention)
- Automation fault

Both of those have to occur at the same time for there to be an accident. But when there is no automation, it is just dependent on driver fault (we know drivers don't pay attention all the time even when they don't have any automation).

So, it all depends on how much less attention drivers would pay if there is automation compared to how much automation helps in masking driver faults. It isn't simple at all.

 

[DarkandStormy]

Who has been arguing that there’s no accidents on FSD beta?

Multiple people on this site.

 

[DarkandStormy]

But that goal doesn't apply to HW3. The goal Elon stated for HW3 is 2-3x safer than a human.

Cyber Rodeo at Giga Texas

00:00 - Pre-show11:15 - Drone Light Show19:12 - Cyber Rodeo Keynote47:38 - Fireworks

www.youtube.com

https://twitter.com/x/status/1430266644971069446

 

[diplomat33]

https://twitter.com/x/status/1650487777912389634

And yes, I believe FSD beta can do this maneuver as well. I just thought it would be interesting to show an example of another system that can also do this maneuver.

 

[diplomat33]

https://twitter.com/x/status/1651223185914839049

 

[diplomat33]

Mobileye released their Q1 financial results.

Summary:

• Revenue increased 16% year over year to $458 million in the first quarter.
• Diluted EPS (GAAP) was $(0.10) and Adjusted Diluted EPS (Non-GAAP) was $0.14 in the first quarter.
• Average System Price increased to $53.9 in the first quarter of 2023 from $51.0 in the prior year period, driven primarily by more than doubling the volume of SuperVision™ sales.
• Additional near-term SuperVision launches include Polestar 4 expected in Q4 2023, and expansion of Zeekr 001 into Europe, expected in late 2023. Multiple development activities with traditional OEMs on SuperVision and Chauffeur™ gaining traction.
• Generated net cash from operating activities of $171 million in the first quarter. Our balance sheet is strong with $1.2 billion of cash and cash equivalents and zero debt as of April 1, 2023.

Mobileye Discloses First Quarter 2023 Results, Updates 2023 Guidance and Provides Business Update | Mobileye

Revenue increased 16% year over year to $458 million in the first quarter. Diluted EPS (GAAP) was $(0.10) and Adjusted Diluted EPS (Non-GAAP) was $0.14 in the first quarter. Average System Price 1 increased to $53.9 in the first quarter of 2023 from $51.0 in the prior year period, driven primarily

ir.mobileye.com

 

[nativewolf]

In context, I meant it is meaningless for removing driver supervision (ie L4 autonomy). It is obviously not meaningless as a stat for the safety of a L2 driver assist system.

Having said that I still think the stat is very misleading. Tesla only measures accidents with airbag deployment or other active restraint employed. So Tesla is only measuring serious crashes. But they measure that accident rate against the overall industry average which includes all police reported accidents and vehicles of different years, some with ADAS and some without. So it is not an apples to apples comparison. Additionally, city driving is at lower speeds so we would expect less of a chance of a crash involving airbag or active restraint deployment, especially if the driver is attentive and taking over before FSD gets into trouble or not using FSD Beta in risky situations in the first place. So the stat is very stacked IMO.

Really, the only thing that the stat says is that an advanced driver assist with an attentive driver will have fewer serious crashes than the average of all accidents. Well, duh, that is not surprising.

Thanks , I realized a couple of years ago that Tesla was using terrible misleading stats on the accident rates. They are essentially useless.

 

[nativewolf]

Multiple people on this site.

Really mostly people on the investment thread

 

[Bladerskb]

I missed this...

https://twitter.com/x/status/1643032816462987264

@EVNow @willow_hiller @powertoold @ZeApelido

comments?

 

[diplomat33]

I missed this...

https://twitter.com/x/status/1643032816462987264

@EVNow @willow_hiller @powertoold @ZeApelido

comments?

Interesting comparison. I wonder why he says that Tesla's FSD has better perception than Mobileye's SuperVision. Honestly, I never really thought that Mobileye's perception was worse than Tesla's. My guess is that it might be based on the FSD Beta visualizations. Tesla has put a lot of work into the FSD Beta visualizations on the screen which do show a lot of objects and is pretty sharp and stable. And the Mobileye visualizations do seem kinda "jumpy". So if you compare the FSD Beta visualizations to Mobileye's FSD visualizations, FSD beta visualizations do look a lot better. And Tesla has spend nearly all their resources for the past few years on just solving vision-only perception. So I guess it should not be that surprising that Tesla's perception might be better than the competition since they devoted so much resources to it. But I think that spending all their resources on just perception has put Tesla behind in other areas like prediction, planning and driving policy. Perhaps that is why he says that Mobileye's Supervision maneuvers better and is more assertive than Tesla's FSD beta. Mobileye has spend resources on HD maps and RSS which help with better maneuvering and assertiveness.

I feel like in the end, Tesla FSD and Mobileye SuperVision may end up being pretty equal since Tesla can catch up on the driving policy part. And if Tesla does decide to fully leverage the fleet to crowdsource HD maps, that will also level the playing field with Mobileye. I think the real key will be making that leap to L4. I think it still remains to be seen if Tesla can get their vision-only to reliable L4. But I think Mobileye will have an advantage jumping to L4 with their HD maps and with their "true redundancy" approach.

 

[powertoold]

That's great and all, but where is Mbleye's equivalent of Tesla's traffic control feature. I thought it should have been released by now.

 

[kabin]

Mobileye released their Q1 financial results.

Summary:

• Revenue increased 16% year over year to $458 million in the first quarter.
• Diluted EPS (GAAP) was $(0.10) and Adjusted Diluted EPS (Non-GAAP) was $0.14 in the first quarter.
• Average System Price increased to $53.9 in the first quarter of 2023 from $51.0 in the prior year period, driven primarily by more than doubling the volume of SuperVision™ sales.
• Additional near-term SuperVision launches include Polestar 4 expected in Q4 2023, and expansion of Zeekr 001 into Europe, expected in late 2023. Multiple development activities with traditional OEMs on SuperVision and Chauffeur™ gaining traction.
• Generated net cash from operating activities of $171 million in the first quarter. Our balance sheet is strong with $1.2 billion of cash and cash equivalents and zero debt as of April 1, 2023.

Mobileye Discloses First Quarter 2023 Results, Updates 2023 Guidance and Provides Business Update | Mobileye

Revenue increased 16% year over year to $458 million in the first quarter. Diluted EPS (GAAP) was $(0.10) and Adjusted Diluted EPS (Non-GAAP) was $0.14 in the first quarter. Average System Price 1 increased to $53.9 in the first quarter of 2023 from $51.0 in the prior year period, driven primarily

ir.mobileye.com

Apparently China EV demand is down as China ended a decade long EV purchase subsidy forcing manufacturers to lower prices, costs, and possibly cut features.

 

[rbt123]

Apparently China EV demand is down as China ended a decade long EV purchase subsidy forcing manufacturers to lower prices, costs, and possibly cut features.

Kinda. The Chinese Association of Automotive Manufacturers is expecting a much slower growth curve but still about 25% more EVs sold this year than in 2022 (8 million vs 5.9 million respectively). Year-over-year growth had been over 50%. They're also buying more cheaper domestic and fewer imported EVs.

Also noteworthy: "This represents 29% of all light vehicles sold in the region, up from 15% in 2021".

China Sees EV Sales Slumping To 8 Million In 2023 | OilPrice.com

China's passenger vehicle sales remained stagnant in March 2023 with a 13.4% fall in sales for the first three months from a year prior, but new energy vehicles showed promise.

oilprice.com

 

[EVNow]

I missed this...

https://twitter.com/x/status/1643032816462987264

@EVNow @willow_hiller @powertoold @ZeApelido

comments?

Hey what happened to ethics ? Are you going to explain your conflict of interests or not ?

 

[EVNow]

Mobileye released their Q1 financial results.

Looks like Tesla price cuts that hit OEMs like Zeekr is having a big impact on MobilEye.

Mobileye Stock Tanks After Earnings. Blame Tesla.

Self-driving-technology firm Mobileye reported strong first-quarter earnings. But its disappointing outlook was a problem.

www.marketwatch.com

 

[Bladerskb]

Interesting comparison. I wonder why he says that Tesla's FSD has better perception than Mobileye's SuperVision. Honestly, I never really thought that Mobileye's perception was worse than Tesla's.

I would hope so seeing as he was test driving SuperVision Gen 1 on the Zeekr 001 that tops out at 32 Tops.
There is SuperVision Gen 2 that uses more power (68 Tops).

Looks like Tesla price cuts that hit OEMs like Zeekr is having a big impact on MobilEye.

Mobileye Stock Tanks After Earnings. Blame Tesla.

Self-driving-technology firm Mobileye reported strong first-quarter earnings. But its disappointing outlook was a problem.

www.marketwatch.com

I mean yeah when you are a tech supplier, units sold by your clients determines how much you make.
According to Tesla fans, Tesla was supposed to supply FSD to every automaker, however we are yet to see one sign up.
How come?