Waymo

[Doggydogworld]

Here is a video from a self-driving car engineer who analyzes JJ's video of the Waymo getting stuck in the construction zone. I figure it might be helpful to hear from someone who works with self-driving cars in their profession.

TL;DW - Initial failure when faced with a few traffic cones is not ideal, but failures are to be expected. The real screwup is how they handled the failure. They didn't (couldn't?) lock the car down, instead allowing it to keep moving unsafely in traffic while the s/w was in an unstable state. The car communicates poorly to other traffic and the company communicated poorly with the customer (lady was nice, but not on the same page as those communicating with the car). He hates the corporate-speak, especially the 'official statement' but also the remote assistance lady following a flowchart instead of interacting naturally.

I don't disagree much with what he says, he just takes way too long to say it. He has a couple of iffy ideas, e.g. letting JJRicks take over and drive the car (he admits this would cause liability issues).

 

[ThomasD]

Didn't Waymo have to rescue the car in a parking lot as well?

 

[JJRicks]

Didn't Waymo have to rescue the car in a parking lot as well?

This one

(yes I used the same thumbnail picture in both videos, don't @ me)

 

[EVNow]

You know what's interesting about this headline ?

Watch self-driving car struggle to understand traffic cones

Doesn't mention the company name. Anyone wants to bet whether "Tesla" would have not been mentioned if this was a Tesla ?

 

[Bladerskb]

You know what's interesting about this headline ?

Watch self-driving car struggle to understand traffic cones

Doesn't mention the company name. Anyone wants to bet whether "Tesla" would have not been mentioned if this was a Tesla ?

that's one article out of the dozens that mention waymo... come on man.

waymo - Google Search

www.google.com

 

[EVNow]

that's one article out of the dozens that mention waymo... come on man.

waymo - Google Search

www.google.com

I'm looking for that one "bullet proof" article ... just as you are looking for that bullet proof EV before considering the harmful effects of your gas car.

 

[Soda Popinski]

that's one article out of the dozens that mention waymo... come on man.

waymo - Google Search

www.google.com

Yeah, not mentioning Waymo seems to be the minority. Instead of searching "Waymo" which will invariably bring up "Waymo" I searched for "self driving car cones" and 5 of 6 mentioned "Waymo" in the headline, with just the La Crosse Tribune not.

 

[diplomat33]

Waymo is actually getting a lot of negative PR from the "cone incident". Pretty much every news outlet is reporting Waymo's failure and mentioning Waymo by name.

 

[diplomat33]

You know what's interesting about this headline ?

Watch self-driving car struggle to understand traffic cones

Doesn't mention the company name. Anyone wants to bet whether "Tesla" would have not been mentioned if this was a Tesla ?

Maybe they did not put Waymo in the headline but Waymo is mentioned in the description right below the video. They are not hiding that it is Waymo.

And pretty much every news outlet is reporting Waymo's failure. Waymo is getting lots of bad PR from the incident. The media is not covering for Waymo.

 

[diplomat33]

Smooth ride!

00:00 Ride start
00:08 Being a rebel
00:23 Rider support call (they're on to me now, oh boy)
00:49 Stop sign left turn
01:20 Unprotected right
02:05 Three way stop
03:25 Unprotected right
04:55 Unprotected right
05:25 Left lane change
06:50 Protected left
11:15 Unprotected left, little bit tricky
11:20 Two right lane changes
12:30 Four way stop
13:13 Waymo can't pronounce Español

 

[diplomat33]

New Waymo presentation:

0:00 Welcoming
1:47 Intro
3:45 Waymo's Mission & History
20:00 The Waymo Driver

At 20 mn, is one of best descriptions of how the 5th gen Waymo Driver works IMO.

 

[EVNow]

Maybe they did not put Waymo in the headline but Waymo is mentioned in the description right below the video. They are not hiding that it is Waymo.

And pretty much every news outlet is reporting Waymo's failure. Waymo is getting lots of bad PR from the incident. The media is not covering for Waymo.

Nothing about cover-up. Its all about clicks - Tesla generates clicks. Hardly anyone has heard of Waymo or Alphabet. Google could (legally) protest if anyone writes the FSD car is Google's.

This is exactly the reason for Alphabet's corporate setup - to protect Google's brand name.

 

[Bladerskb]

Nothing about cover-up. Its all about clicks - Tesla generates clicks. Hardly anyone has heard of Waymo or Alphabet. Google could (legally) protest if anyone writes the FSD car is Google's.

This is exactly the reason for Alphabet's corporate setup - to protect Google's brand name.

So 99% of articles mentions Waymo in the title but the one article that doesn’t proves your point? Lol

 

[powertoold]

Waymo's been ramping up their gen-5 i-Pace cars in San Francisco big time. I saw more than 5 of them near the Dolores Park area within the span of 15 minutes yesterday. They all seemed to be going the same route though.

 

[diplomat33]

Waymo's been ramping up their gen-5 i-Pace cars in San Francisco big time. I saw more than 5 of them near the Dolores Park area within the span of 15 minutes yesterday. They all seemed to be going the same route though.

It might have something to do with Waymo applying for a permit to charge customers for ride-hailing in SF. It would make sense that they might be ramping up testing in preparation for launching the commercial ride-hailing service when they get approved. In any case it is exciting to see Waymo doing more with the 5th Gen FSD. Thanks for sharing.

 

[gangzoom]

Waymos very public failure shows the problem of NNs. They aren't really 'intelligent' so fails quite spectacularly when faced with new road situations which humans can easily solve.

Getting stranded on a dual carriageway could have ended very badly, especially with buses having to swerve to avoid the car.

The process wasn't vision or HD maps, but the software. Since Tesla is using the same NN approach, why do we think Tesla will do better with FSD??

 

[mikes_fsd]

Waymos very public failure shows the problem of NNs. They aren't really 'intelligent' so fails quite spectacularly when faced with new road situations which humans can easily solve.

Waymo's failure is not a problem of NNs!
It demonstrates - in flying colors - the sunk cost fallacy and inability to make massive changes when the need arises.
They - Waymo - just double down on stupid!

 

[DanCar]

Waymos very public failure shows the problem of NNs.

Do you think Waymo can turn it around? What would it take for Waymo failure to be successful? What if they announced soon they are offering service in SF? Would that make them seem like less of failure?

 

[Microterf]

Do you think Waymo can turn it around? What would it take for Waymo failure to be successful? What if they announced soon they are offering service in SF? Would that make them seem like less of failure?

I think when they deploy in SF (and how much of SF they cover) is going to determine a lot for them as a company. If it goes smooth, it will be great, and will show they can scale quickly. If it doesn't go well, they'll have a hard time convincing external investors to give them more money

 

[mikes_fsd]

Do you think Waymo can turn it around? What would it take for Waymo failure to be successful? What if they announced soon they are offering service in SF? Would that make them seem like less of failure?

I think Waymo could show that they are able to scale by showing the following:

• Expanding to new metro areas (say starting with 50 sq miles in each new metro)
• Showing the time it takes for them to go from the 50 sq mile limit to full city public roads.
• How quickly they can expand to new metros (if they can add a new top 25 city every quarter - there is potential to actually make some money)
• Expansion not dependent on expanding remote operators (and eliminating shadow vans entirely)
  • if Chandler AZ takes 10 remote operators today expanding to another city (say El Paso) should not double the remote operators but maybe 10% increase?
  • obviously they will need fleet management in each city, so those are not part of these limits
  • but shadow cars with ppl that follow driverless taxi's should not be a thing, only on demand calls when needed.
• Just deploying to SF is not enough if they triple they staff in every department.