Waymo

[Bitdepth]

You are just interjecting in a banter you don't understand.

I can read, you aren't writing a foreign language that no one can decipher. You are trying to equate miles driven by a L4 ADS vs miles driven by a L2 ADAS

To which @AlanSubie4Life corrects you by saying when FSD is engaged, it is understood and stated by Tesla that the software is not driving but "ASSISTING" therefore it is legally not liable for any accidents. Vs a L4 system that is responsible for driving and is responsible for any accident.

When FSD is engaged, it is driving - irrespective of the legal liability i.e. it is doing the steering, accelerating and braking etc. I'm just supervising - and as soon as I disengage, I'm driving .... We are talking about who is controlling the car, not who is legally liable.

We are talking about who is responsible for driving as stated by the manufacturer of the system and their design intent for the system not wishful thinking by a careless user. One is driver assist, a really good one and another is a fully autonomous driver, a really good one.

If you want to continue this stupid argument - you can start a new thread.

You are in here trying to equate miles driven by 2 different system with different design designation (L4 vs L2), if you want to continue this stupid argument, make a thread about ADAS systems and compare those. Don't do it in this thread about a L4 system.

 

[EVNow]

You are trying to equate miles driven by a L4 ADS vs miles driven by a L2 ADAS

No. Re-read.

We are talking about what kind of data Tesla & Waymo can get. I've no idea why you are talking about liability.

When you engage FSD - and disengage because you don't like what FSD is doing, Tesla can get some data. Same as Waymo can get some data if the car hits something. As for as getting data is concerned, when FSD is engaged, FSD is driving - that there is a safety driver or not doesn't matter.

 

[Bitdepth]

No. Re-read.

We are talking about what kind of data Tesla & Waymo can get. I've no idea why you are talking about liability.

When you engage FSD - and disengage because you don't like what FSD is doing, Tesla can get some data. Same as Waymo can get some data if the car hits something.

It's not rocket science.

With an average of 5 miles (generous given the small area) per ride, we get 50k per week or about 2.5 Million miles per year vs ~360 Million miles for FSDb (total US/Canada).

BTW, at $1 per mile, that is a grand total of $2.5 M of revenue per year.

You are comparing two different miles. One is L4 ADS taxi service the other is L2 ADAS.

If you want to compare L4 Miles FSDb will be 0 vs 50K miles per week. This is what @AlanSubie4Life pointed out. While in theory you can compare anything, it is not useful for the purposes of discussing miles driven by a robo taxi service. The miles you or anyone has while FSDb is engaged does not equal the miles Cruise, Waymo, Baidu has when operating an autonomous taxi service. You can compare it to Superceuise, Bluecruise, Supervision, NIO NAD and the myriad other ADAS available to consumers to purchase.

Furthermore I think we are well passed the data discussion. Mobileye has the the world's largest data set of any company working on autonomous vehicle. Some 200 petabytes of video clips (self proclaimed) but no other company is claiming otherwise.

 

[EVNow]

It's not rocket science.

And yet you fail

I was responding to this.

Doubtful given that Waymo and Tesla are currently two separate markets. But, I do wonder how 10,000 mi/wk on Waymo compares to the number of miles per week of FSD miles. Not that they are equivalent, but the ratio must now be staggering.

 

[Bitdepth]

And yet you fail

I was responding to this.

Comparison of Waymo miles vs FSDb miles.
Yea its not rocket science.

Doubtful given that Waymo and Tesla are currently two separate markets. But, I do wonder how 10,000 mi/wk on Waymo compares to the number of miles per week of FSD miles. Not that they are equivalent, but the ratio must now be staggering.

 

[EVNow]

Yea its not rocket science.

And yet to fail !

Let me spell it out slooowly for you.

@Supcom wondered how the numbers compare. So, I gave an estimate.

What's your problem ?

 

[Bitdepth]

And yet to fail !

Let me spell it out slooowly for you.

@Supcom wondered how the numbers compare. So, I gave an estimate.

What's your problem ?

Let's spell it out for you in numbers.

FSDb 350 million miles do not equal 2.5 million Waymo miles. There is no statistics that shows Tesla is getting any more from 350 million FSDb miles than Waymo gets from 2.5 million miles.

Waymo went from a prototype in 2009 to a working robo taxi service in 2018 = 12 years
While inventing much of the technology many if not all use today for neural networks.

Tesla goes from an ADAS in 2015 to a more advanced ADAS in 2023 = 9 years.

Those miles do not equal each other.

 

[powertoold]

Whatever the case, Tesla's data collection stack gathers more edge cases than Waymo's, can't put a number on it, but at least one or two orders of magnitude more edge cases.

Tesla aspires to become a generalized robotaxi service with their data collection stack. FSDb works on 4 million miles of USA roads and more in Canada...

I'd be surprised if Waymo drives on more than 100,000 miles of USA roads.

 

[KLam]

Let's spell it out for you in numbers.

FSDb 350 million miles do not equal 2.5 million Waymo miles. There is no statistics that shows Tesla is getting any more from 350 million FSDb miles than Waymo gets from 2.5 million miles.

Waymo went from a prototype in 2009 to a working robo taxi service in 2018 = 12 years
While inventing much of the technology many if not all use today for neural networks.

Tesla goes from an ADAS in 2015 to a more advanced ADAS in 2023 = 9 years.

Those miles do not equal each other.

May I add that most of the FSDb miles are garbage miles with a lot of errors and self introduced edge cases. I bet the Waymo miles are more refined and with less and less errors.

 

[Supcom]

May I add that most of the FSDb miles are garbage miles with a lot of errors and self introduced edge cases. I bet the Waymo miles are more refined and with less and less errors.

From an engineering standpoint, I expect that the most useful data comes from the errors. Those are the situations that need to be improved.

 

[diplomat33]

Map of service area expansion in SF:

 

[EVNow]

Map of service area expansion in SF:

Oh - finally all but the most important commercial areas of SF

 

[diplomat33]

Oh - finally all but the most important commercial areas of SF

SF will add that last part soon. They are almost done with the SF expansion. You know Waymo's approach is to gradually expand the geofence until it eventually covers every important area.

 

[EVNow]

SF will add that last part soon. It is a gradual expansion over time.

With an emphasis on gradual

Critics might say, painfully slow instead of gradual.

 

[Bladerskb]

Oh - finally all but the most important commercial areas of SF

With an emphasis on gradual

Critics might say, painfully slow instead of gradual.

Wait are you complaining about a greyed out 1.6 sq miles area temporarily having no pick-up/ drop off spots because its densely packed with pedestrians and likely has no available parking spots? Which would primarily mean that they didn't want to pick you up or drop you off far from your destination and receive bad PR? An area that's available to employees?

Surely there's far bigger issues to criticize...

 

[EVNow]

Which would primarily mean that they didn't want to pick you up or drop you off far from your destination and receive bad PR?

Where most of Uber's pick ups and drop offs are ?

Its like including all the areas near an airport except for the terminals

 

[Bitdepth]

Where most of Uber's pick ups and drop offs are ?

Its like including all the areas near an airport except for the terminals

It's almost as if this is a hard problem to solve and doing so safely is the number one goal and priority.

12 months ago, Waymo would avoid the entire grey region.

Then 6 months ago they included it but limited only for employees.

And now 6 months later, they have included every part for everyone except they won't do pick-ups and drop offs in a 1 sq mile bordering 3 streets obviously due to congestion.

It's almost like they are testing their technology and making clear steady progress. Who knows what would happen 3 maybe 6 months from now.

 

[EVNow]

It's almost like they are testing their technology and making clear steady progress. Who knows what would happen 3 maybe 6 months from now.

True - its only taken them 13 years. I'm sure they will get to it before my kids born after they started the project start driving cars.

 

[Bladerskb]

True - its only taken them 13 years. I'm sure they will get to it before my kids born after they started the project start driving cars.

How long has it taken Tesla? And yet still nothing...

 

[EVNow]

How long has it taken Tesla? And yet still nothing...

Nothing for ICE lovers like you .... but plenty for people like me.