Autonomous Car Progress

[diplomat33]

I don’t think this has been posted here although it’s a month old. Made me laugh.

It was posted awhile back. There was an entire thread about it.

 

[Dewg]

Not really.

Tesla tells you in writing on the screen of the car it's doing it, and requires you to agree to let them do it, AND it doesn't send Tesla any biometric info either- it processes the video feed in real time in the car, the only info Tesla ever gets is if you had any strikeouts from not paying attention- it doesn't send Tesla anything biometric.

It's a nonsense class action to try and make some lawyers some money.

Ford and Caddy and many others also use in-car cameras for driver attentiveness tracking- but they don't make nearly as big a headline when you sue them I guess.

Hopefully you're right - but with juries you never know.

 

[bradtem]

Here is a much better article with details.

Basically, BMW will become part of the software IP owning partners, rather than just a downstream consumer of Mobileye. Interestingly they retain the right to all the REM map details collected from their Mobileye partnership.

BMW, Qualcomm And Arriver To Collaborate On New Driver Assist And Automation

In November 2021, Qualcomm announced that its new Snapdragon Ride compute platform for advanced driver assist systems (ADS) and partially automated driving (ADS) had been selected for future BMW vehicles.

www.forbes.com

Car OEMs have been in a quandry about this. They correctly worry that the self-driving system is going to be the most important part of the car, that it will be one of the key drivers of the value -- and the core value if the car will be a robotaxi.
They are very afraid of not owning it, just being the company that plugs it into their car. They have gotten more used to that state with the Tier Ones but that's because the Tier Ones happily have no ego and take a back seat. MobilEye is now Intel which is a different class of company. It's the reason they always had trouble doing a deal with Waymo. They don't want to be the Foxconn.

So I think the BMW deal does say there is competition for MobilEye, which there wasn't in the existing Tier One space -- though there is competition in the tech company space. But it also reflects their strong desire to be masters of their own fate.

 

[EVNow]

Car OEMs have been in a quandry about this. They correctly worry that the self-driving system is going to be the most important part of the car, that it will be one of the key drivers of the value -- and the core value if the car will be a robotaxi.
They are very afraid of not owning it, just being the company that plugs it into their car. They have gotten more used to that state with the Tier Ones but that's because the Tier Ones happily have no ego and take a back seat. MobilEye is now Intel which is a different class of company. It's the reason they always had trouble doing a deal with Waymo. They don't want to be the Foxconn.

So I think the BMW deal does say there is competition for MobilEye, which there wasn't in the existing Tier One space -- though there is competition in the tech company space. But it also reflects their strong desire to be masters of their own fate.

Reminds me - I was having lunch with a consultant friend of mine. He was talking about what big auto OEM c-suits talk about during strategy presentations. This was a few years back.

They would almost never mention EVs, they were not afraid of EVs eating their lunch.

But they were very afraid of AVs.

 

[bradtem]

Reminds me - I was having lunch with a consultant friend of mine. He was talking about what big auto OEM c-suits talk about during strategy presentations. This was a few years back.

They would almost never mention EVs, they were not afraid of EVs eating their lunch.

But they were very afraid of AVs.

If they were wise they were afraid of both. One key problem that happens is that by the time a technology is nibbling on your lunch it is too late to stop it from eating all of it later.

They will look and say, "EVs are only 3% of auto sales" and tell themselves there is no big worry there. Well, they would if it weren't for the fact Teslas are outselling their models in many places and Tesla stock is worth more than all of them together. That woke them up. 3% EV sales would not have.

Same for self driving. They will say, "Sure, a few companies have cars in a few cities, this is afar in the future." And indeed, the saturation of the tech is far in the future, but it far too late well before that happens.

They don't think it's too late. They think "we will adapt." They are trying like hell to adapt on EVs but they are not that good at it. They think it's just about replacing the power train.

 

[EVNow]

If they were wise they were afraid of both.

They assumed AVs will happen faster than EVs. Turns out they were wrong.

 

[stopcrazypp]

They assumed AVs will happen faster than EVs. Turns out they were wrong.

Could also be they think the challenge of EVs would be easier to counteract than AVs (given most of them are behind in the software front). I remember in the period before it was sure Tesla would survive or be profitable as a company, the Tesla naysayers largely argue the large automakers can easily wait things out.

Basically they can just wait for new players like Tesla to take all the risks to build up the supply chain and demand, and then they can just swoop in and benefit from lower battery and EV parts prices, while stealing all of Tesla's sales using their own brand recognition. I seem to recall in premium market Audi using this strategy, and in general market VW and Toyota. Of course things, didn't turn out quite so easy for them (and even today demand far outstrips supply for Tesla).

 

[bradtem]

They assumed AVs will happen faster than EVs. Turns out they were wrong.

Not sure if they thought that. I mean, obviously not because EVs have been out for decades, some of them made by the big car OEMs. For a long while Nissan was the EV ruler, but ruled a very tiny kingdom while they all made compliance vehicles. AVs were always going to come later they were not as scared of EVs, they figured, as I said, that EVs were just replacing the power train and making cars just the same as they always did, an evolutionary change which they would clearly win.

AVs on the other hand were scarier. They were not obvious winners. High tech companies were leading the way. And they were the replacement for the car, the thing that comes after the car. That's much more frightening than cars with a new motor, even if the EVs came first.

Tesla caught them off guard. They didn't just replace the motor, they rethought many parts of the car. They made it a software device. They were not ready for that.

 

[Jeff N]

It was posted awhile back. There was an entire thread about it.

Yep, just found it. Thanks. Somehow I missed it earlier.

A Satire (Downfall Hitler Rant) on the Tesla FSD team trying to tell the boss that FSD has problems

I saw you liked my more serious videos. This one is in fun. And everybody who has seen one of the thousands of other Hitler videos should know they are not in any way trying to compare the subject with Hitler, just do some satire. Enjoy.

teslamotorsclub.com

 

[diplomat33]

@bradtem has a nice video on various news stories that relate to autonomous cars.

 

[thesmokingman]

The MIT video below is about their cheetah robot. What's interesting is that it they tried a different approach this time. They allow the neural net to run in a virtual environment in which they can speed up time which bypasses the very long learning process. It's pretty interesting to see the difference in the way the robot runs. That got me wondering what if the same idea was somehow applied to autonomous driving? I dunno the technicals but the idea of it is pretty cool.

Want to make robots run faster? Try letting AI take control

We’re still pushing the limits of legged robots.

www.theverge.com

By using a simulator, engineers can place the robot in any number of virtual environments — from solid pavement to slippery rubble — and let it work things out for itself. Indeed, the MIT group says its simulator was able to speed through 100 days’ worth of staggering, walking, and running in just three hours of real time.

 

[bradtem]

The MIT video below is about their cheetah robot. What's interesting is that it they tried a different approach this time. They allow the neural net to run in a virtual environment in which they can speed up time which bypasses the very long learning process. It's pretty interesting to see the difference in the way the robot runs. That got me wondering what if the same idea was somehow applied to autonomous driving? I dunno the technicals but the idea of it is pretty cool.

Want to make robots run faster? Try letting AI take control

We’re still pushing the limits of legged robots.

www.theverge.com

This is the approach many teams are taking (to varying degrees.) Wayve is 100% neural network. Tesla is constantly trying to grow how much they use. Waymo still keeps things in layers but makes heavy use of machine learning at all layers. Almost nobody isn't doing it.

 

[Jeff N]

I just got back from an 18-minute 4-mile drive in San Francisco in the back seat of a Cruise Bolt EV without safety drivers. This was a free trip as part of the recently begun nighttime public robotaxi test service.

I’m still figuring out exactly how much I can say but I don’t think I’m violating any rules in saying that it handled it like a champ.

Much of it was on long stretches just stopping for stoplights in mild traffic but there were several times with cars parked in the middle of the road on residential streets with emergency lights blinking that it deftly handled. The last stretch included some curvy very narrow hilly roads with many cars parked along the curb. I’ll give it an A+.

For comparison, I’m going to try recreating the route on my Y with waypoints, if necessary (likely), and then will drive it with 10.11 at a similar time of the evening within the next few days.

 

[voyager]

I'll keep saying that there is a HUGE difference with driverless during the daytime when vehicles and people are crowding the streets
in congested inner cities. And of course under less than ideal weather conditions.

 

[bradtem]

I'll keep saying that there is a HUGE difference with driverless during the daytime when vehicles and people are crowding the streets
in congested inner cities. And of course under less than ideal weather conditions.

It is more difficult. Cruise does it in the day with safety drivers who almost never intervene, but not yet never, so they are not willing to do it empty. Waymo does it empty in the day, in the easier town of Chandler. AutoX does it empty in the day, in the more complex suburbs of Shenzhen.

I would never trust my Tesla to do it anywhere, night or day, by a longshot.

 

[EVNow]

I would never trust my Tesla to do it anywhere, night or day, by a longshot.

Ofcourse you shouldn't - why would you even think of that ?

 

[Daniel in SD]

Ofcourse you shouldn't - why would you even think of that ?

Because the software is supposed to be safer than a human by the end of the year?

 

[EVNow]

Because the software is supposed to be safer than a human by the end of the year?

*checks date* .... come back on 9 months

 

[Daniel in SD]

*checks date* .... come back on 9 months

It just seems ridiculous to question why someone would think about literally the most important metric of a full self-driving system while testing it.

 

[EVNow]

It just seems ridiculous to question why someone would think about literally the most important metric of a full self-driving system while testing it.

Because *no one* is claiming it is an unsupervised system *now*.

Its like trying to use the current SpaceX rocket to fly to Mars .... after all that's the objective, right ?

Afterall, Waymo won't run in my neighborhood ...