Autonomous Car Progress

[diplomat33]

Some interesting moments in this Cruise driverless ride:

0:00 Preview
0:09 Start of Ride
2:09 Nudging for open door
4:34 Awkward trajectory crossing Market Street
4:48 Stopping for unknown
5:10 Slow to proceed at stop sign
5:46 Pull over
6:09 Stuck at pull out
6:27 Car attempts to call support
7:07 Car switches gears to park, then reverse
7:14 Reversing with remote assistance
7:42 Call disconnects with no representative
10:09 Squeezing past a lead vehicle at a right turn
12:01 Unprotected left turn
12:28 No one picks up the phone
13:38 Giving up right of way at stop sign

 

[EVNow]

Some interesting moments in this Cruise driverless ride:

0:00 Preview
0:09 Start of Ride
2:09 Nudging for open door
4:34 Awkward trajectory crossing Market Street
4:48 Stopping for unknown
5:10 Slow to proceed at stop sign
5:46 Pull over
6:09 Stuck at pull out
6:27 Car attempts to call support
7:07 Car switches gears to park, then reverse
7:14 Reversing with remote assistance
7:42 Call disconnects with no representative
10:09 Squeezing past a lead vehicle at a right turn
12:01 Unprotected left turn
12:28 No one picks up the phone
13:38 Giving up right of way at stop sign

I'm starting to wonder if Cruise is really much better than FSD. Their disengagement rates don't seem to be right - they need to report every intervention.

 

[diplomat33]

I'm starting to wonder if Cruise is really much better than FSD. Their disengagement rates don't seem to be right - they need to report every intervention.

I agree. They should report all remote assistance events, Waymo too.

They don't report remote assistance events as a disengagement since remote assistance does not disengage the autonomous driving (remote assistance provides a hint but the autonomous driving stays on the whole time). So technically, a remote assistance event is not a disengagement. This allows them to report a very good disengagement rate, even though they have a bunch of remote assistance events. This does skewer the public perception of their autonomous driving.

Personally, from the videos I've seen, I think Cruise is probably slightly better than FSDbeta but not by much. Cruise does seem to make a lot "FSD Beta" type mistakes IMO.

 

[EVNow]

This allows them to report a very good disengagement rate, even though they have a bunch of remote assistance events. This does skewer the public perception of their autonomous driving.

They should be asked to report every intervention - because that's a failure. I mean most of the FSD YouTubers report number of interventions ... !

 

[diplomat33]

They should be asked to report every intervention - because that's a failure. I mean most of the FSD YouTubers report number of interventions ... !

I agree. Remote assistance events should count as an intervention.

And I really think intervention reports should come from Tesla, not youtubers. Youtubers reporting interventions is not a statistically accurate measurement.

 

[Bladerskb]

I'm starting to wonder if Cruise is really much better than FSD. Their disengagement rates don't seem to be right - they need to report every intervention.

They should be asked to report every intervention - because that's a failure. I mean most of the FSD YouTubers report number of interventions ... !

How come you don't demand that of Tesla?

Isn't it curious how you treat these companies differently? You take Tesla at their word, accepting the numbers they present without a hint of skepticism, despite them providing 1-2 sentence statements. Contrast this with Waymo, which has shared hundreds of pages of detailed data and analysis. Yet, you seemingly deem it insufficient.

The narrative becomes even more perplexing as you are now scrutinizing Cruise. Calling them a liar. Suspecting deception.
The double standards is incredible.

 

[EVNow]

How come you don't demand that of Tesla?

Isn't it curious how you treat these companies differently?

Because they are supposed to be robotaxis and Tesla is an ADAS system?

 

[Bladerskb]

Whether a manufacturer is willing to license tech from a major competitor has little to do with if it was superior. For example, it's pretty well acknowledged that superchargers are the superior charging option in the US, yet no major automaker has been willing to license it. There's a multitude of possible reasons brought up through the years:
- automakers balking at upfront infrastructure investments Elon suggested was required for licensing/access to stations
- hubris and branding issues from using a Tesla station
- Tesla's patent terms being unfavorable to larger manufacturers with a larger patent pool than Tesla (something Tesla only recently tried to address by opening up the standard)
- Actively helping Tesla, a major competitor, get better

These issues are avoided by choosing a more neutral third party (which CCS is the DC charging equivalent).

Notice the details of what you pointed out
1 - Tesla Superchargers are universally viewed as superior
2 - OEM would have to pay huge costs to use (including non-standard charging cable)
3 - Provides little to no benefit to OEM and their customers
4 - Government *were* planning on giving billions towards EV infrastructure

So guess what? They go the alternative route which is to support Electrify America (the public standard) while also working on their own so they can get the government handouts.
Almost all EV manufacturers have their own dedicated charging infrastructure.
This is happening both in the US, EU and China.

So this is not compatible at all with the FSD /AV / ADAS situation.

Again I remind you that it is the Tesla thought leaders that are pushing the "Only Tesla will have FSD".
"Tesla will have a monopoly on FSD", "Everyone will be forced to license from Tesla". "Tesla is 10 years ahead"... etc

If there were any truth to that. Even just a 0.1% truth to it. Then OEMs would be flocking to Tesla FSD.
Because Governments are not planning on funding ADAS/AV development and there are tangible benefits to the OEM and customers from AV (robotaxi) and ADAS, etc.

But yet we have nothing. No one. Nada. Is even looking at Tesla FSD let alone wanting to use it.
You can't explain this away by saying "But superchargers..."

 

[EVNow]

And I really think intervention reports should come from Tesla, not youtubers. Youtubers reporting interventions is not a statistically accurate measurement.

I agree - but they are not going to do it until it improves a lot.

 

[Jeff N]

Some interesting moments in this Cruise driverless ride:

Wow, that was a huge customer support fail. All of my in-car support calls or car-initiated calls were answered within 10-15 seconds.

I have seen the backing up issue before. In one case, the car stopped to pick me up too close to a parked car ahead at the curb. When it came time to start the trip, the car waited for ~20 seconds and then began nudging forward ever so gingerly for another 15 seconds before finally driving off at a normal speed. It just narrowly avoided needing to call for remote assistance.

 

[powertoold]

I'm starting to wonder if Cruise is really much better than FSD. Their disengagement rates don't seem to be right - they need to report every intervention.

It's always been clear to me. The AV companies simply run their trains on the easy mode tracks in loops to game their disengagement stats. If you ran fsd beta between the two easiest destinations during low traffic hours, how many miles would you get between disengagements?

That's why I say that fsd beta offers the only pure way to assess autonomous driving progress right now. And why I hate the levels. Why I dislike any marketing material from AV companies, etc. Etc.

 

[stopcrazypp]

Notice the details of what you pointed out
1 - Tesla Superchargers are universally viewed as superior
2 - OEM would have to pay huge costs to use (including non-standard charging cable)
3 - Provides little to no benefit to OEM and their customers
4 - Government *were* planning on giving billions towards EV infrastructure

So guess what? They go the alternative route which is to support Electrify America (the public standard) while also working on their own so they can get the government handouts.
Almost all EV manufacturers have their own dedicated charging infrastructure.
This is happening both in the US, EU and China.

So this is not compatible at all with the FSD /AV / ADAS situation.

Again I remind you that it is the Tesla thought leaders that are pushing the "Only Tesla will have FSD".
"Tesla will have a monopoly on FSD", "Everyone will be forced to license from Tesla". "Tesla is 10 years ahead"... etc

If there were any truth to that. Even just a 0.1% truth to it. Then OEMs would be flocking to Tesla FSD.
Because Governments are not planning on funding ADAS/AV development and there are tangible benefits to the OEM and customers from AV (robotaxi) and ADAS, etc.

But yet we have nothing. No one. Nada. Is even looking at Tesla FSD let alone wanting to use it.
You can't explain this away by saying "But superchargers..."

I'm pointing out superchargers simply because it is a clear example that a technology that was superior doesn't mean that automakers will license it. That the specific details might be different is irrelevant. It shows that there is a lot more that goes into an automaker deciding to license a technology than if it was superior. As such, just because no automaker has opted to license it is not an valid argument that it's not superior (you obviously can point out other metrics though that might demonstrate it as inferior or superior).

If you want a closer example instead of talking about Tesla, you can look at automakers moving away from Mobileye (which is generally considered the superior option for non-Tesla ADAS). They are doing so because they want more control over the software, regardless if Mobileye currently is technically superior than other offerings.
Ford’s Lane Change Signals a Shift Away From AV Development | The Ojo-Yoshida Report

 

[Bladerskb]

I'm pointing out superchargers simply because it is a clear example that a technology that was superior doesn't mean that automakers will license it. That the specific details might be different is irrelevant. It shows that there is a lot more that goes into an automaker deciding to license a technology than if it was superior. As such, just because no automaker has opted to license it is not an valid argument that it's not superior (you obviously point out other metrics though that might demonstrate it as inferior or superior).

But that's the point, the specific details around ADAS & AV is not a negative for licensing FSD.
If FSD was truly 10 years ahead and are months away from L5, then OEMs would make billions from licensing Tesla's FSD.

On the other hand, supercharger's details were a negative for OEMs even though it was superior. By not choosing the supercharger they don't have to invest huge money upfront, they can use public charging standard like EA and still be able to setup their own charging network and get free millions from the government funding. So there were strong reasons why they shouldn't use superchargers. I just don't see that with Tesla FSD.

If you want a closer example instead of talking about Tesla, you can look at automakers moving away from Mobileye (which is generally considered the superior option for non-Tesla ADAS). They are doing so because they want more control over the software, regardless if Mobileye currently is technically superior than other offerings.
Ford’s Lane Change Signals a Shift Away From AV Development | The Ojo-Yoshida Report

Mobileye is a great example because when the shift from basic ADAS to high-end ADAS / AV started, OEMs had to make a decision whether to outsource or go in-house. It's all about the pros/cons to each OEM. However OEMs who decides to outsource is going to Mobileye, Huawei, etc. Remember it's not automatic that these OEMs pick Mobileye. Mobileye still has to go out, present to them and convince them in order to win the contract. So for the companies that want to outsource, why isn't Tesla winning the contracts?

Recent contractual wins for Mobileye are from Polestar (Geely), another recently is Porche/Audi (VW). Both either used or planned to use in-house systems and switched to Mobileye. It's worth pointing out that Mobileye still has a ~70-75% overall market share.

Another potential contractual win that is up in the air and will be announced this year is either Ford or Stellantis based on what Mobileye have said in their Investors Call. So the EyeQ5 wins started out flat but are now picking up. So it's clear that these contracts are up for grabs yet Tesla isn't winning them. Why?

 

[powertoold]

Take it from a dude who's been testing all sorts of EV brands lately:

 

[stopcrazypp]

But that's the point, the specific details around ADAS & AV is not a negative for licensing FSD.
If FSD was truly 10 years ahead and are months away from L5, then OEMs would make billions from licensing Tesla's FSD.

On the other hand, supercharger's details were a negative for OEMs even though it was superior. By not choosing the supercharger they don't have to invest huge money upfront, they can use public charging standard like EA and still be able to setup their own charging network and get free millions from the government funding. So there were strong reasons why they shouldn't use superchargers. I just don't see that with Tesla FSD.

Many of the same things would still apply to FSD:
- The initial theoretical terms Elon may offer for third party usage of FSD may be just as unfavorable as it was for superchargers (meaning a substantial upfront lump payment of some sort).
- Using a Tesla solution is just as awkward for branding.
- Tesla still has the unfavorable patent policy unless Tesla decides to make FSD an open standard like they eventually made the TPC connector (which there is zero indication they have any plans to do so).
- Using a Tesla solution helps Tesla improve more than it does for the other automaker

Mobileye is a great example because when the shift from basic ADAS to high-end ADAS / AV started, OEMs had to make a decision whether to outsource or go in-house. It's all about the pros/cons to each OEM. However OEMs who decides to outsource is going to Mobileye, Huawei, etc. Remember it's not automatic that these OEMs pick Mobileye. Mobileye still has to go out, present to them and convince them in order to win the contract. So for the companies that want to outsource, why isn't Tesla winning the contracts?

Recent contractual wins for Mobileye are from Polestar (Geely), another recently is Porche/Audi (VW). Both either used or planned to use in-house systems and switched to Mobileye. It's worth pointing out that Mobileye still has a ~70-75% overall market share.

It's troubling however that the companies with the systems generally considered the best non-Tesla systems (Super Cruise, and the current improved Blue Cruise) are moving away from Mobileye (BMW also is moving away from Mobileye, but their implementation doesn't currently stand out much, so it's less of a surprise there). Polestar and VW's systems are not generally considered that good, so I'm not sure that is much of a win (edit: and it looks like VW is looking to move to Qualcomm also the long run for future platforms, with Mobileye still being an interim solution)
VW eyes quicker uptake of Qualcomm chips in key platforms

It shows that just because a supplier has the superior system, doesn't mean an automaker would stick with them. It remains to be seen how well the Qualcolmm examples from GM and Ford will do vs their previous Mobileye based systems.

Another potential contractual win that is up in the air and will be announced this year is either Ford or Stellantis based on what Mobileye have said in their Investors Call. So the EyeQ5 wins started out flat but are now picking up. So it's clear that these contracts are up for grabs yet Tesla isn't winning them. Why?

Tesla isn't even offering FSD contracts. The only thing Tesla ever did was Elon did a tweet about the possibility of offering FSD to other manufacturers. Then nothing came of it (no more details, not rumors of talks, unlike for superchargers where Elon did at least give some detail to the terms of third party usage of superchargers). I think the idea of third parties not contracting to FSD as an indication of quality is a red herring.

 

[Bladerskb]

Many of the same things would still apply to FSD:
- The initial theoretical terms Elon may offer for third party usage of FSD may be just as unfavorable as it was for superchargers (meaning a substantial upfront lump payment of some sort).

But the payoff would be yuuuuge as they would essentially be getting a L5 robotaxi in return.

- Using a Tesla solution is just as awkward for branding.

Not at all, unlike the Tesla supercharger that can't be white-labeled. Tesla's FSD can actually be white-labeled.
This is what companies do all the time. This is for example what Mobileye does.
The software will be called whatever the OEM wants, it will have its own UI and presentation, etc.
For example Mobileye Supervision on the Zeekr 001 is called the Zeekr Pilot.
BlueCruise on other cars are called ActiveGlide.
Honda licenses SuperCruise from GM and calls it something else.

- Tesla still has the unfavorable patent policy unless Tesla decides to make FSD an open standard like they eventually made the TPC connector (which there is zero indication they have any plans to do so).

I don't see how this applies here

- Using a Tesla solution helps Tesla improve more than it does for the other automaker

This is simply not true as Tesla already has enough cars with their system.
What's that 4+ million cars with HW3? But OfCourse for the tesla fans, the narrative don't stop.
They will continue pushing it, "Tesla will soon get 1 trillion miles of data each year, no one else is close".
The stupidity knows no bounds. Ignorance begats Ignorance.

It's troubling however that the companies with the systems generally considered the best non-Tesla systems (Super Cruise, and the current improved Blue Cruise) are moving away from Mobileye (BMW also is moving away from Mobileye, but their implementation doesn't currently stand out much, so it's less of a surprise there).

Its actually not surprising because its all about the value proposition that Mobileye is offering in exchange for what they can get somewhere else or what they can do for themselves. GM owns Cruise and also has a good inhouse team that created Ultra Cruise. Ford hired 550 Algo AI engineers and formed a new team (LATITUDE AI) to work on L3+ systems.

The value proposition that Tesla would offer them is billions of dollars of profits which Mobileye is clearly not offering.

Polestar and VW's systems are not generally considered that good, so I'm not sure that is much of a win (edit: and it looks like VW is looking to move to Qualcomm also the long run for future platforms, with Mobileye still being an interim solution)
VW eyes quicker uptake of Qualcomm chips in key platforms

But that's the point, Polestar and VW were looking for a solution and didn't pick Tesla FSD.

It shows that just because a supplier has the superior system, doesn't mean an automaker would stick with them. It remains to be seen how well the Qualcolmm examples from GM and Ford will do vs their previous Mobileye based systems.

Because of the value proposition. GM for example felt they could get better value proposition by running everything in-house and it has clearly worked out for them as they are in final stages to release Ultra Cruise. But the value proposition that Tesla would offer them would be billions of dollars of profits which Mobileye is clearly not offering. If we are to believe what Elon and Tesla fans are saying.

Tesla isn't even offering FSD contracts. The only thing Tesla ever did was Elon did a tweet about the possibility of offering FSD to other manufacturers. Then nothing came of it (no more details, not rumors of talks, unlike for superchargers where Elon did at least give some detail to the terms of third party usage of superchargers). I think the idea of third parties not contracting to FSD as an indication of quality is a red herring.

How would you know they are not offering contracts? Elon has mentioned it several times on different Earnings Calls.
If Tesla were having talks you wouldn't know about it because talks like that are kept secret. You didn't hear rumors of Polestar or Audi/Porche talks that Mobileye were having (even though i knew well ahead of time). Majority of talks are not leaked.

In conclusion, Either Ford or Stellantis is planning to license Mobileye's Supervision. In Tesla's own backyard. How is that possible?
You are claiming its possible because Tesla isn't offering any contracts?

 

[Jeff N]

Uh oh…..

Fortunately, Cruise support was on the job quickly. I received an automatic phone support call within seconds and an on-scene Cruise support vehicle was there immediately. That’s them in the right side of the windshield parked in a regular Bolt with the orange Cruise logo on rear panel.

Actually, it almost looks like we collided with them.

And, actually, they were already there when my car pulled up at its planned stopping point. The “collision” happened just a few feet earlier as a pedestrian deliberately walked up close to the car as it was about to make the turn into the side street where it was planning to drop me off.

The collision was a false alarm.

Film at 11 (or sooner if I can get a good network signal).

 

[Jeff N]

My first collision.

 

[diplomat33]

https://twitter.com/x/status/1660677337036517376

 

[EVNow]

https://twitter.com/x/status/1660677337036517376

I think at some point the regulators will step in ... they can't have "driverless" drives so many interventions ...

Or rather, regulators *should* step in. But GM has bought them off, so they won't.