Autonomous Car Progress

[nativewolf]

Yes - I understand the current system is setup for trucks. In the long run - its better to optimize the system for trains. Electrified trains is the most climate friendly way to move goods.

Well I am not sure of that in the USA. The cost of building roads and/or trains has to take into account the huge losses associated with the infrastructure. Really solving rail issues means a huge amount of additional rail and rail yards and the losses of land associated with them.

In light of the massive changes going on battery capabilities I would hope we disperse with lots of road/rail needs and go for more direct air freight on EV airframes. That might take 20 years to play out but them we get to keep a lot more high value ag land in ag. Unless we actually learn to make tunnels at scale, if tunnels are possible at scale we could start to unwind a lot of craptastic urban infrastructure.

 

[JB47394]

In light of the massive changes going on battery capabilities I would hope we disperse with lots of road/rail needs and go for more direct air freight on EV airframes.

Check through this report by the Bureau of Transportation Statistics. You might reconsider wanting to try to solve any large scale freight problems with air transportation.

Moving Goods in the United States

Freight Facts and Figures - Moving Goods in the United States

data.bts.gov

 

[nativewolf]

Check through this report by the Bureau of Transportation Statistics. You might reconsider wanting to try to solve any large scale freight problems with air transportation.

Moving Goods in the United States

Freight Facts and Figures - Moving Goods in the United States

data.bts.gov

I am part of the problem, being in forestry/logging. We aspire to be a really decent sized problem child. The best we can do is electric trucks, in fact it would revolutionize the logistics side.

 

[EVNow]

Well I am not sure of that in the USA. The cost of building roads and/or trains has to take into account the huge losses associated with the infrastructure. Really solving rail issues means a huge amount of additional rail and rail yards and the losses of land associated with them.

At one point this exact same point was made for not supporting EVs.

There is a lot of land available for freight trains - and tracks. Just need upgrades and massive public infrastructure investment like they do in China (and India as well). Apparently all we want to do is tax cuts and defense.

 

[stopcrazypp]

At one point this exact same point was made for not supporting EVs.

There is a lot of land available for freight trains - and tracks. Just need upgrades and massive public infrastructure investment like they do in China (and India as well). Apparently all we want to do is tax cuts and defense.

Analogy doesn't quite work. EV infrastructure is drastically easier to build, far more flexible (does not have to be at an exact location), and brings potential benefits to the site in terms of business traffic (a railroad track however brings almost all negatives: noise, looks bad aesthetically, causes inconvenience and anger for crossing to other side).

Those kind of issues causes a strong NIMBY reaction, which is why large scale rail projects have a hard time (see California's HSR, which has a largely supportive population, but yet it is not going well). Countries like China can do it, but they don't have to worry about NIMBY (anyone in the way has no say).

 

[diplomat33]

 

[spacecoin]

May Mobility goes driverless (EA) in Sun City (NW of Phoenix), Arizona with fixed stops.

May Mobility launches first driverless transit service for Early…

May Mobility today announced the launch of the company’s first driverless service for riders on public roads in Sun City, Arizona.

maymobility.com

"Through May Mobility’s first rider-only service, a select group of Early Riders in Sun City will have the ability to request a pickup in one of May Mobility’s Toyota Sienna Autono-MaaS vehicles from a variety of stops. The rider-only service will initially operate on public roads Monday through Friday in the afternoon and will be available at no cost to Early Riders."

 

[spacecoin]

https://twitter.com/x/status/1736890863358505307

 

[diplomat33]

Nvidia talk on end-to-end architecture for next gen AVs:

 

[diplomat33]

BOO! Introducing Wayve Ghost Gym, a closed-loop neural simulator where we can train, test and debug our end-to-end AI driving models.Come and take a look inside https://bit.ly/3GPRm1U

https://twitter.com/x/status/1737760300563071079

 

[diplomat33]

More info on what Mobileye will share at CES 2024:

Mobileye at CES 2024 launch and keynotes

Mobileye at CES 2024

Prof. Shashua will also introduce a new technological breakthrough that unlocks automakers’ ability to configure the driving experience in ways that align with their brand and consumer tastes.

• Mobileye: Now. Next. Beyond. with President & CEO Prof. Amnon Shashua, on January 9, 2024, at 11:00 a.m. PT. (11:00 a.m. - 12:00 p.m. PT)

As a follow up to Now. Next. Beyond., Mobileye CTO Prof. Shai Shalev-Shwartz will present:

• Mobileye's Driving Experience Platform: Architecture, Abstractions, and APIs on Wednesday, January 10 at 1:00 p.m. PT. (1:00 -1:40 p.m. PT)

This talk will delve into Mobileye's specialized programming platform engineered specifically to meet this challenge. Prof. Shalev-Shwartz will showcase how the platform's APIs provide automakers with the tools to program the AI's functionality to the unique desired driver characteristics of each car model.

Mobileye Announces CES 2024 Press Conference with Prof. Amnon Shashua | Mobileye News

Also on Wednesday, January 10, Mobileye Senior Vice President of AV Johann “JJ” Jungwirth will appear on the CES-presented Conference Session panel titled The Middle Lane: Self-Driving Cars Today at 9:00 a.m. PT.

https://www.reddit.com/r/SelfDrivingCars/comments/18nlr0o

Calling it a tech breakthrough seems like a bit of a stretch. But I am looking forward to it. Hopefully, we get some good info on how it will work and what it will mean for consumers. Hopefully, we also get some info on when Mobileye SuperVision and Chauffeur will be deployed and to what vehicle models and when. And I can see how being able to customize the automated driving experience will be great for carmakers but as a consumer I am a bit concerned because it means that two vehicles both with the same Mobileye system might have different driving experiences. So as a consumer, I might not know what driving experience to expect. I know Shashua recently talked about how the perception stack will be the same and the "hard" planning will be the same, and carmakers will only get to customize the controls, "soft" planning, driver monitoring and UI. But I still wonder especially since some carmakers might implement the tech better than others.

 

[diplomat33]

https://twitter.com/x/status/1737944921732972643

 

[diplomat33]

https://twitter.com/x/status/1739017055263674670

 

[KArnold]

Max speed 37 MPH. Need to start somewhere.

 

[diplomat33]

Xpeng expands XNGP to 52 cities with plans to cover 200 cities in China next year:

https://twitter.com/x/status/1740735113204474085

https://twitter.com/x/status/1740722993402462607

 

[kabin]

Xpeng expands XNGP to 52 cities with plans to cover 200 cities in China next year:

https://twitter.com/x/status/1740735113204474085

It appears to be about as good as FSD as it brakes while sharply steering into the direction of a near miss at the 45sec mark. It will be interesting to see how the Chinese culture/government responds after a few accidents/deaths.

 

[DanCar]

Flowpilot is comma AI / openpilot branched.

https://twitter.com/x/status/1740777052989800513

GitHub - flowdriveai/flowpilot: flow-pilot is an openpilot based driver assistance system that runs on linux, windows and android powered machines.

flow-pilot is an openpilot based driver assistance system that runs on linux, windows and android powered machines. - flowdriveai/flowpilot

github.com

 

[scottf200]

Xpeng expands XNGP to 52 cities with plans to cover 200 cities in China next year:

https://twitter.com/x/status/1740735113204474085

https://twitter.com/x/status/1740722993402462607

I really can't wait for Tesla to get into China and then when those comparisons happen with XNGP and others we will see how good FSDbeta is.

https://twitter.com/x/status/1740951640901927255

 

[stopcrazypp]

I really can't wait for Tesla to get into China and then when those comparisons happen with XNGP and others we will see how good FSDbeta is.

https://twitter.com/x/status/1740951640901927255

I would be surprised if FSD Beta ends up better, given it's so fitted to US conventions and roads. Tesla also currently uses a more inferior navigation map provider in China last I looked. I don't think that necessarily predicts the performance if both were on equal footing. Probably releases in Europe will be more telling (there are some Chinese makes selling in Europe).

 

[scottf200]

I would be surprised if FSD Beta ends up better, given it's so fitted to US conventions and roads. Tesla also currently uses a more inferior navigation map provider in China last I looked. I don't think that necessarily predicts the performance if both were on equal footing. Probably releases in Europe will be more telling (there are some Chinese makes selling in Europe).

Excellent points about the majority source of training data likely being USA-centric. You are right that Europe seems more equal ground.
BTW, I do/did not think that FSDbeta would be better. I don't think the cameras are sufficient. I'm fairly happy with FSDbeta but it makes so many simple mistakes and is so inconsistent. Such interesting times tho.