Waymo

[diplomat33]

Interesting negotiation here where the other car seemed to have the right of way but it yields to the Waymo.

 

[Bitdepth]

Also, the little details like notifying your passenger of oncoming cyclists when dropping them off so they don't open the door and cause an accident.

I feel like people are underestimating what is required to create a safe autonomous Taxi service.

 

[diplomat33]

https://twitter.com/x/status/1636400282295730177

 

[Bitdepth]

What happens when a road is blocked and waymo has to obey a temporary stop sign.

 

[diplomat33]

What happens when a road is blocked and waymo has to obey a temporary stop sign.

I feel this video illustrates why Waymo is ahead of Cruise. Simply put, Cruise planning stack does not seem to be as good. It seems to get confused more. On the other hand, Waymo has much more advanced perception/prediction/planning stack. Waymo can read temporary signs but more importantly, understands the scene and does not get confused. Not to say that Waymo can never get confused, but I think Waymo gets confused less often than Cruise.

 

[EVNow]

I wonder how Baidu and other Chinese players compare to Waymo.

Anyone have an inkling?

 

[diplomat33]

I wonder how Baidu and other Chinese players compare to Waymo.

Anyone have an inkling?

Well, Baidu just got their permit to give driverless rides in Beijing. This makes their 3rd city, after Wuhan and Chongqinq. Baidu says they will start with 10 driverless cars in Beijing's Yizhuang economic development zone.

Baidu's self-driving ride-hailing service platform Apollo Go has won a permit to offer fully driverless rides in Beijing, allowing it to expand the service to three Chinese megacities.

The search engine giant announced the development today, saying it is the first provider of fully driverless robotaxi services in the capital city of any country worldwide. Apollo Go has previously been approved to offer the service in Wuhan and Chongqing.

Baidu Apollo will deploy a total of 10 fully driverless vehicles in Beijing's Yizhuang Economic Development Zone, according to a press release from the company.

Source: Baidu's robotaxi platform Apollo Go gets permit to offer fully driverless rides in Beijing

 

[EVNow]

Well, Baidu just got their permit to give driverless rides in Beijing. This makes their 3rd city, after Wuhan and Chongqinq. Baidu says they will start with 10 driverless cars in Beijing's Yizhuang economic development zone.

Yes - I saw that. Infact my question was triggerred by this report.

We don't know how "good" the driving is. The Chinese website @Bladerskb linked to once has mostly smaller drives (atleast of XPeng) - wonder whether there are longer independent drive videos to make a comparison.

 

[diplomat33]

We don't know how "good" the driving is. The Chinese website @Bladerskb linked to once has mostly smaller drives (atleast of XPeng) - wonder whether there are longer independent drive videos to make a comparison.

I did find this video of DeepRoute.AI driving autonomously for 1 hour in China (but video is sped up). It has some commentary and FSD visualization.

But yeah, it is hard to say since I don't know of any long videos from Baidu's driverless.

 

[Bladerskb]

This is what I mean by Tesla is simply behind. It still makes simple planning mistakes that leads to immediate collisions. A Waymo or Cruise CANNOT do that. IF they did it would be automatic accident as there would be no driver.

Here you see two accidents that were prevented by the driver. Do you think that Waymo would crash in this scenario? Serious question?
There are worse scenarios than this in SF. The negotiating ability has nothing to do with geofence.

43:06 - Near Accident
28:16 - Almost Hits a Truck

Waymo's driving in heavy rain is impressive, much better than I was expecting. The spazzy steering wheel still gives me anxiety though.

What I find interesting about this is that in the FSD videos we see that small variations in individual junctions, weather, the behaviour of other vehicles, local variations in road markings and driving practice etc make significant differences in how well FSD behaves.

But for some reason we see one video of Waymo handling a given situation well and the default assumption is that they can do that just as well everywhere. Seems rather inconsistent.

Only if you disregard geofencing and obvious segmentation of the market.

It is very clear from last 5 years that geo-scaling is exceedingly difficult for traditional AV companies.

I think the jury is out on who is "ahead". I also think the question is moot - neither have shown the ability to be really at level 5 (i.e. can operate in all ODD).

I'd say Waymo is ahead in geofenced robotaxi market and Tesla is ahead in consumer market.

In that case I propose the Disneyland monorail is the best. As I understand it that’s been fully autonomous for years if not decades, with a proven track record. They’ve nailed it - all they need to do is roll it out worldwide and they’ve cracked autonomous transit.

An exaggeration to illustrate the point, obviously, but the point is that the horizontal scaling is in and of itself part of the challenge. Saying that Waymo are close to fully autonomous driving in the particular regions they’ve developed their system on is valid, but I think it’s underestimating the challenges presented by different geographies to say they’re further ahead than Tesla in the universal sense.

Tesla seem to have built a learning system that will be able to scale since they have automated a lot of the data capture, labelling and training processes and they have a fleet in the wild to capture test data and perform testing on a vast scale compared to Waymo. They’ve got a challenge in that they’re working with a less capable sensor suite and they don’t have a good ability to pivot on previous hardware design decisions because of the deployed fleet.

 

[EVNow]

This is what I mean by Tesla is simply behind.

Dude, don't quote me for no reason. I don't quote you everytime GM is "behind" or Cruise stops in the middle of nowhere.

 

[diplomat33]

What I find interesting about this is that in the FSD videos we see that small variations in individual junctions, weather, the behaviour of other vehicles, local variations in road markings and driving practice etc make significant differences in how well FSD behaves.

But for some reason we see one video of Waymo handling a given situation well and the default assumption is that they can do that just as well everywhere. Seems rather inconsistent.

It is not inconsistent. We are not saying Waymo can handle all similar situations perfectly based on a video. That is clearly not the case since Waymo has instances of "stalls" or needing remote assistance. But we can assume that Waymo can handle all safety critical cases with acceptable safety. We can assume this not because of a video, but because it is driverless. By removing the safety driver, Waymo is essentially self-certifying that the car meets their standard of "safe enough". Tesla has not removed the safety driver so Tesla has not self-certified that FSD Beta is safe enough yet. If Tesla did remove the safety driver then we could make the same assumption as with Waymo that it can handle all safety critical cases with acceptable safety.

Typically, I think a safety critical case is defined as a collision resulting in serious damage or injury.

Here is a handy reminder:

1) If the autonomous driving cannot handle all safety critical cases then it needs a safety driver.
2) If the autonomous driving can handle all safety critical cases on its own but not all non-safety critical cases, then it can be driverless but needs remote assistance.
3) If the autonomous driving can handle all cases both safety critical and non-safety critical on its own then it does not need a safety driver or remote assistance.

Nobody has reached #3 but companies like Waymo have reached #2 within a geofence hence why they are driverless but have remote assistance.

 

[Bladerskb]

Dude, don't quote me for no reason. I don't quote you everytime GM is "behind" or Cruise stops in the middle of nowhere.

Its better to try to stop and reach a minimal risk condition than to ram into other cars. I'm not saying that Cruise is doing a great job at it.
But its a big difference. You simply CAN'T hit other cars/objects.

My point still stands, if you take a Waymo and wanted to deploy it driverless into Chicago 50 sq mile area for a 24/7 service...How long will it take Waymo? I would say currently 4 months. This is the difference between Tesla and Waymo. For Tesla its indefinite, for Waymo its just acouple months of testing.

Do you disagree?

 

[Mullermn]

Nobody has reached #3 but companies like Waymo have reached #2 within a geofence hence why they are driverless but have remote assistance.

I note you include reference to the geofence there, quite correctly. Most of this thread seems to disregard that when it’s convenient though, and talk as if working within the geofence is some sort of guarantee that it works everywhere.

Take for example the post above - Tesla is ‘simply behind’ because there are videos of an FSD car messing up on YouTube. Completely (and quite significantly) ignoring the difference in approach between the two companies and implying that a Waymo would have dealt with that scenario despite it being entirely unproven.

Insert standard disclaimer - not dragging Waymo, just think that the vastly different aims and methodologies of the two companies are intrinsic to any evaluation of how successful they are being.

 

[EVNow]

Do you disagree?

Yes I disagree - Tesla and Waymo are not in the same market. Tesla sells the best ADAS there is to consumers for a fat profit. Waymo loses billions running 2 driverless trips a day in San Francisco.

Come back here and tag me when Waymo stuff is running in consumer cars or Waymo is operating at the scale Uber does (which also loses billions !).

When your employer releases FSDb competitor in consumer cars- that would be a good comparison.

 

[aronth5]

This is the classic it "depends". In some situations Waymo is ahead and in some Tesla is ahead. There is plenty of room for multiple vendors with different approaches.

 

[EVNow]

This is the classic it "depends". In some situations Waymo is ahead and in some Tesla is ahead. There is plenty of room for multiple vendors with different approaches.

The problem is some haters only want to see the best of Waymo and worst of Tesla.

 

[Mullermn]

The problem is some haters only want to see the best of Waymo and worst of Tesla.

There are those who are just as biased the other way. The curious thing about this forum (not this thread particularly), given that is nominally for Tesla enthusiasts, is how many Tesla detractors there are here.

 

[JHCCAZ]

There are those who are just as biased the other way. The curious thing about this forum (not this thread particularly), given that is nominally for Tesla enthusiasts, is how many Tesla detractors there are here.

The biggest problem I find is that whatever I post, if not clearly identifiable as being in the Love or Hate camp, seems often regarded with confusion or suspicion. Taking a position somewhere between, or at least not being simply along the one-dimensional spectrum, turns out to be not so easy.

I suspect there are plenty of members and readers who can appreciate those intermediate positions, but one has to be careful not to be drawn into endless tangents of I'm rRght How Can You [Fanboys | Haters] Be So Wrong.

Regarding Waymo and CruIse at. al., I can praise some of their engineering and deployment without endorsing the entire Robotaxi futurist paradigm, and also without regarding them as direct competition for what I want to do with my Tesla - which is to have it be my personal car that can one day drive me around independently. The latter is something that is not presently in the business model or cost consideration of waymo, and unfortunately seems to be getting less endorsement from Ford, GM and the other major incumbents. (I think it is in the business model of Mobileye, but for whatever reasons the automakers seem to be holding them at arm's length.)

 

[Bladerskb]

Yes I disagree - Tesla and Waymo are not in the same market. Tesla sells the best ADAS there is to consumers for a fat profit. Waymo loses billions running 2 driverless trips a day in San Francisco.

Come back here and tag me when Waymo stuff is running in consumer cars or Waymo is operating at the scale Uber does (which also loses billions !).

When your employer releases FSDb competitor in consumer cars- that would be a good comparison.

This makes no sense. It baffles my mind that when discussing and comparing technologies, people turn the discussion into marketing and economics.
This is like saying you can't compare two nba players (MJ, Lebron for example) because they play different positions, and one makes more money than the other.

Again we are talking about Technological capabilities and reliability for autonomous driving. Turning this into a marketing and economics argument is basically you resigning that Tesla is technologically behind in autonomous driving.

Rather than comparing performance of both technologies in similar road environments and conditions (Downtown Phoenix, SF, LA, Chicago, etc).
You instead comeback with "Tesla makes more money than Waymo LMFAOOOO"

Like what.....?

Again this video of Waymo handling a deadly wind storm in SF further breaks the myth that continues to be fostered and propelled by Tesla propagandists like Andrej Karpathy. I remember when he was saying something to the effect of how every leaf on a tree has to be mapped and if something moves the whole thing falls apart.

Yet here we are in a huge wind storm. Not perfect, but that's some Disney world train right? @Mullermn