Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Waymo

This site may earn commission on affiliate links.
You mean he didn't admit / say, there were no customers? So all that stuff about buying so many Porsche (?) cars was a bluff?

What makes people believe whatever other company folks say but not Tesla ? I look at all company communication as marketing.
I don't think anyone believes AV company timelines anymore... At least I hope not!
 
  • Like
Reactions: mark95476 and EVNow
As a FSDbeta users, I certainly like how Tesla is releasing and deploying software builds every ~2 weeks.

These other AV companies are seemingly trying to counter by spinning more marketing spiel and press releases.

The other AV companies also deploy new software updates on a regular basis. It's just that we the public don't see them because the updates are internal, to a fleet with employee safety testers instead of public beta testers.
 
I did not say Waymo's updates are every 2 weeks, just that they are regular.

In any case, it's great that Tesla is doing FSD Beta every 2 weeks. I like getting frequent updates.
Yes, when the release is internal - really doesn't matter whether the releases are every hour or once a year.

What matters are the FSD Beta releases.

But definitely there are some "interested parties" in the media / competitors (is there a difference ?) who try to portray 2-week releases as evil.
 
I don't know if there are newer blogs about mapping from Waymo. But this a year old blog is on their site ...


Nowhere do they claim the mapping and labelling is entirely automated, like some of the folks here do.
 
  • Like
Reactions: sfuape
I don't know if there are newer blogs about mapping from Waymo. But this a year old blog is on their site ...


Nowhere do they claim the mapping and labelling is entirely automated, like some of the folks here do.

Correct. Waymo says in the blog that "most" of the map updating is automated. So it is highly automated but not fully automated.

We’ve automated most of that process to ensure it’s efficient and scalable.

But in context, it is talking about updating the HD map, not the original creation. But we can assume that the creation part is similarly highly automated, but maybe not completely automated.
 
But in context, it is talking about updating the HD map, not the original creation. But we can assume that the creation part is similarly highly automated, but maybe not completely automated.
I saw that - but didn't quote it because it only referred to some updates.

Since a lot of people apparently are very confident that the whole mapping & labelling is "fully automated", surely there are better links stating so ?
 
Here is the abstract of Waymo's paper:

3D object detection is a key module in safety-critical robotics applications such as autonomous driving. For such applications, we care the most about how the detections impact the ego-agent’s behavior and safety (the egocentric perspective). Intuitively, we seek more accurate descriptions of object geometry when it’s more likely to interfere with the ego-agent’s motion trajectory. However, current detection metrics, based on box Intersection-over-Union (IoU), are object-centric and are not designed to capture the spatio-temporal relationship between objects and the ego-agent. To address this issue, we propose a new egocentric measure to evaluate 3D object detection: Support Distance Error (SDE). Our analysis based on SDE reveals that the egocentric detection quality is bounded by the coarse geometry of the bounding boxes. Given the insight that SDE can be improved by more accurate geometry descriptions, we propose to represent objects as amodal contours, specifically amodal star-shaped polygons, and devise a simple model, StarPoly, to predict such contours. Our experiments on the large-scale Waymo Open Dataset show that SDE better reflects the impact of detection quality on the ego-agent’s safety compared to IoU; and the estimated contours from StarPoly consistently improve the egocentric detection quality over recent 3D object detectors.

Full paper: https://openreview.net/pdf?id=OMNRFw1fX3a

It sounds like Waymo found a way to make 3D object detection better.
 
It sounds like Waymo found a way to make 3D object detection better.

I'm glad they're talking about this topic. So often you see folks assume that LIDAR automatically equals 3D object detection, but it's just a cloud of data points. It looks impressive because you can show people a visualization of the data points and they go "Wow! I can see all the details in that object over there!" but that's taking human-level intuition to composite that point-cloud into distinct 3D objects.
 
Boris Sofmam, head of trucking at Waymo, says that they've fundamentally changed the hardware and software stack from what's deployed in Phoenix because the 4th generation system was not easily scalable.

Of course each version is more scalable, more accurate, less expensive, etc. They said they same thing about Gen 4 and will say it about Gen 6, Gen 7, etc.

Gen 4 was never designed to scale to millions or even 100s of thousands. They knew they'd have Gen 5 by then. But Gen 4 was designed to scale to 10s of thousands. Their words and actions show it.

You don't stagnate your fleet at 5-10 operational cars because your tech can't scale to millions. You do stagnate your fleet if you realize you can't attract enough customers to make a viable business.
 
  • Like
Reactions: EVNow
You don't stagnate your fleet at 5-10 operational cars because your tech can't scale to millions. You do stagnate your fleet if you realize you can't attract enough customers to make a viable business.
Its funny how Waymo scaling reflects another area Google promised a lot and delivered little - Google Fiber - but I guess for totally different business reasons.
 
  • Like
Reactions: mark95476
Waymo has given hundreds of people robotaxi rides since its test rollout in San Francisco in August, with tens of thousands more residents on a waitlist, the Alphabet Inc (GOOGL.O) company's co-chief executive said on Friday at the Reuters Next conference.


Tens of thousands of people are on the waitlist to use Waymo in SF! So the demand is clearly there. As soon as Waymo is sure the autonomous driving is safe in SF, I hope they deploy a much bigger fleet than they have in Chandler to tap into that demand.
 
I hope they deploy a much bigger fleet than they have in Chandler to tap into that demand.
Even bigger than the 5-10 cars deployed in Chandler? WoW!
(I know they have more cars there for R&D, but the support guy said the Waymo One service itself only uses 5-10).

Unlike Phoenix, though, there are definitely plenty of customers assuming they can solve the more difficult technical and operational problems. Here's hoping it doesn't take them too long.
 
  • Like
Reactions: diplomat33
Tens of thousands of people are on the waitlist to use Waymo in SF! So the demand is clearly there. As soon as Waymo is sure the autonomous driving is safe in SF, I hope they deploy a much bigger fleet than they have in Chandler to tap into that demand.
Ofcourse there will be a lot of people in SF. Its the nerd-capital of the world ;)

Besides, very difficult to park - that means a lot of people come there in mass transit / uber. May be not that many now with a lot of offices still closed ... but by the time Waymo is ready probably even Covid may be over !