GM's Cruise tweets cryptic message, plans announcement on January 21

[willow_hiller]

Here's the tweet: Cruise on Twitter

Those coordinates correspond to:

• Dahuting Han Dynasty Tombs
• Merthyr Tydfil, UK
• Automuseum Dr. Carl Benz
• Thomas Edison Center at Menlo Park
• Foundation Guglielmo Marconi
• UCLA
• Google/PARC
• Midtown Manhattan
  
• Cruise HQ

Speculations?

 

[diplomat33]

Here's the tweet: Cruise on Twitter

View attachment 500809

Those coordinates correspond to:

• Dahuting Han Dynasty Tombs
• Merthyr Tydfil, UK
• Automuseum Dr. Carl Benz
• Thomas Edison Center at Menlo Park
• Foundation Guglielmo Marconi
• UCLA
• Google/PARC
• Midtown Manhattan
  
• Cruise HQ

Speculations?

My guess is some kind of robotaxi demo in those different areas to show that their autonomous driving can work in very diverse driving environments. Maybe some type of cross country, coast to coast, robotaxi demo?

"time to move beyond the car" could be a reference to robotaxis. We know Cruise wants to have a true driverless car with no steering wheel or pedals. GM even showed a Bolt with no steering wheel or pedals.

 

[willow_hiller]

"time to move beyond the car" could be a reference to robotaxis. We know Cruise wants to have a true driverless car with no steering wheel or pedals. GM even showed a Bolt with no steering wheel or pedals.

Even without a steering wheel or pedals, I would still consider a Bolt to be a "car." But what would qualify at that point? Autonomous bus? Autonomous airplane?

 

[CrayZ1]

"Beyond the car" - Some other use for autonomy software.

I'll admit the coordinates baffle me. The first three are essentially nowhere unless they are just pointing at a country.
The letters spell out something?

 

[diplomat33]

Even without a steering wheel or pedals, I would still consider a Bolt to be a "car." But what would qualify at that point? Autonomous bus? Autonomous airplane?

They could just be referencing how we use cars and the paradigm shift that a robotaxi represents. But yeah, they could be referencing something else like an autonomous "flying car". That would certainly be "moving beyond the car" too.

 

[Mardak]

I'll admit the coordinates baffle me. The first three are essentially nowhere unless they are just pointing at a country.

The coordinates are of places relating to various inventions in transportation (e.g., wheels, rails, engines) and communication (e.g., phones, radio, internet). So the last coordinate is Cruise HQ suggesting they've made an equally important invention probably involving both transportation and commnication.

 

[diplomat33]

The coordinates are of places relating to various inventions in transportation (e.g., wheels, rails, engines) and technology (e.g., radio, internet, AI). So the last coordinate is Cruise HQ suggesting they've made an equally important invention.

Good catch. I do think the "important invention" they made is a true driverless car.

 

[outdoors]

Rumor was that a while back there were looking at moving on from here.com on the mapping for Cruise. They were looking at using Google mapping for the Cruise. Not sure this would be anything with those dates.

 

[Daniel in SD]

They better be deploying robotaxis in San Francisco, otherwise this is super lame.
Or maybe they're pivoting to autonomous flying cars which would probably be a smart move since the useless autonomous vehicle market is completely saturated right now

 

[S4WRXTTCS]

My guess is that they're talking about moving beyond car ownership. Their CEO is really opposed to car ownership.

We Need to Move Beyond the Car

So I think it's likely going to be a Robotaxi announcement of sorts. Maybe they'll mix in V2X communication stuff.

 

[diplomat33]

Cruise just posted two videos of their autonomous driving:

30 minutes of continuous autonomous city driving:

1 hour of continuous autonomous city driving:

 

[Daniel in SD]

Cruise just posted two videos of their autonomous driving:

Still waiting for that 1 million mile video!
It would be much more interesting to see videos of failures. I wonder what problems they're still having. Hopefully this all means they feel that they're ready to deploy their vehicles without safety drivers.

 

[Daniel in SD]

1 hour of continuous autonomous city driving:

They really need to have their software avoid getting caught behind people turning left!

 

[diplomat33]

Still waiting for that 1 million mile video!
It would be much more interesting to see videos of failures. I wonder what problems they're still having. Hopefully this all means they feel that they're ready to deploy their vehicles without safety drivers.

If they are ready to deploy, then presumably, they are not experiencing any serious failures, at least not any safety critical failures.

 

[willow_hiller]

1 hour of continuous autonomous city driving:

Starting at around 4:26 in the "Hour Continuously" video, I'm really impressed how the vehicle takes two speed bumps. It's hard to see, but if you slow down the video they reduce speed from 20 MPH to about 10 MPH before hitting each bump.

Since I've got speed bumps near my house, I've been thinking for a while that camera-only vehicles may have trouble seeing speed-bumps when they're the same color as the road. I wonder if LIDAR is able to pick them up, or whether they're mapped.

 

[diplomat33]

Starting at around 4:26 in the "Hour Continuously" video, I'm really impressed how the vehicle takes two speed bumps. It's hard to see, but if you slow down the video they reduce speed from 20 MPH to about 10 MPH before hitting each bump.

Since I've got speed bumps near my house, I've been thinking for a while that camera-only vehicles may have trouble seeing speed-bumps when they're the same color as the road. I wonder if LIDAR is able to pick them up, or whether they're mapped.

Not mapped, most likely LIDAR sees the speed bumps. I don't think LIDAR would have any trouble with speed bumps. LIDAR is an active sensor so it will simply get a reflection off the speed bump that will tell the car the shape and location of the speed bump. Cameras are more difficult because you have to train the vision to interpret the pixels as a speed bump and figure out how far away the speed bump is. If the speed bumps are the same color as the road, it will blend in with the road, making it harder for the camera to see them. So yes, cameras will have a harder time with speed bumps. On the other hand, LIDAR just sends out a pulse and measures the time of the reflection and calculates the distance.

This is one of many examples why Tesla's method is harder. It can be done with vision but it is harder because you have to train the NN and you can miss edge cases like speed bumps that are the same color as the road. But with LIDAR, it's relatively easier because it just measures the reflection of a pulse.

 

[Bladerskb]

Another video

 

[S4WRXTTCS]

If they are ready to deploy, then presumably, they are not experiencing any serious failures, at least not any safety critical failures.

My favorite thing about Cruise is they picked a city where lots of people hate driving in.

I don't even drive in SF.

Screw that.

 

[diplomat33]

Kyle Vogt, Co-founder and CTO of Cruise just posted a blog on Medium:
The Disengagement Myth

In the blog, he argues that disengagement rates alone are a poor metric for measuring autonomous driving because not all disengagements are the same and the conditions of the drives are not the same. He believes we need hard empirical evidence than an AV is better than humans in the same environment and we need to look at the rate of change in the performance of the autonomous car to see if the AV is getting better. He also argues that short demos can be too easily curated to only show the best parts so he argues for long unedited videos like the ones Cruise just released. He also teases that more details will be shared at the Jan 21 event.

I am definitely more excited for the Jan 21 event. Looks like Cruise will have some very positive progress to share.

 

[motocoder]

They’re falling back to the horse and buggy. It is GM, after all...