Autonomous Car Progress

[EVNow]

What city do u live in if u don’t mind me asking?

Eastern suburb of Seattle - with a high concentration of Google, Amzn & MSFT employees

ps : Chances of me moving to a different place before Waymo (if ever) gets here is high.

IMO, Waymo will have AVs in your city way before the 2040's.

How do you know ? It may fold before that ...

ps : From 2019 ...

Tesla has freeway NOA that works fairly well. Waymo doesn't work in my city.

I bet we'll get City NOA on Tesla before Waymo works in my city.

 

[diplomat33]

Eastern suburb of Seattle - with a high concentration of Google, Amzn & MSFT employees

ps : Chances of me moving to a different place before Waymo (if ever) gets here is high.


How do you know ? It may fold before that ...

Seattle is a pretty big city with lots of Google employees as you point out. I imagine Seattle is probably on waymo's top 10 cities to expand to. So i think waymo could expand to Seattle as early as 2025-2030. Waymo will.come to your city way before it comes to mine.

Sure waymo could fold by then but I highly doubt it. Waymo has lots of capital and Waymo is making great progress and starting to scale. They have a clear path to scaling. By early 2024, waymo will likely have fully driverless in chandler, downtown phoenix, san Francisco, LA and highway. And with the progress they are making with ML, waymo could solve the 4 domains (dense urban, suburb, highway and weather) in 5 years, maybe sooner. I read all of waymo's ML research, it is state of the art. After that, it is profits and scaling everywhere. IMO, there is very little chance waymo goes under.

 

[EVNow]

Seattle is a pretty big city with lots of Google employees as you point out. I imagine Seattle is probably on waymo's top 10 cities to expand to. So i think waymo could expand to Seattle as early as 2025-2030. Waymo will.come to your city way before it comes to mine.

I'm in an eastern suburb ... so, not sure it will come to us that quickly.

Also, they have to get really comfortable driving in rain ...

ps : BTW, has anyone seen how well Waymos handle roundabouts ?

 

[kabin]

ps : BTW, has anyone seen how well Waymos handle roundabouts ?

Waymo does well on a relatively busy but simple single lane roundabout. No signs of indecisiveness or steering stutter. You might want to turn the volume off/down.

Waymo roundabout

 

[EVNow]

Waymo does well on a relatively busy but simple single lane roundabout. No signs of indecisiveness or steering stutter. You might want to turn the volume off/down.

Waymo roundabout

Would be interesting see how it handles a busy multi-lane roundabout.

 

[diplomat33]

I'm in an eastern suburb ... so, not sure it will come to us that quickly.

It is unknown when Waymo will come to the eastern suburb of Seattle of course. I am just offering my educated guess. Waymo already handles suburbs well so that part should not be a technical challenge to Waymo.

Also, they have to get really comfortable driving in rain ...

Yes. They are working on that. Waymo CEO said that they can handle rain, wet roads and puddles safely and smoothly. But I am sure there is more testing to be done. Like you said, they need to get "really comfortable". But they are making progress. I am confident they will get really comfortable with rain in a year.

 

[EVNow]

It is unknown when Waymo will come to the eastern suburb of Seattle of course. I am just offering my educated guess. Waymo already handles suburbs well so that part should not be a technical challenge to Waymo.

But they need to be able to connect the city to suburbs using very congested freeways. BTW, have they started working on freeways ?

BTW - I don't think it is actually very difficult for Waymo to cover a city. They need to focus on the city, map it, test thoroughly and then be able to cover it. Just that there are so many cities ... that is where scaling comes. How fast can you scale ....

 

[diplomat33]

BTW, have they started working on freeways ?

Yes. They announced accelerated testing on freeways in September.

https://twitter.com/x/status/1565384559830732803

BTW - I don't think it is actually very difficult for Waymo to cover a city. They need to focus on the city test thoroughly and be able to cover it. Just that there are so many cities ... that is where scaling comes. How fast can you scale ....

Yeah scaling is the big question. Everybody wants to scale as fast as possible. Waymo's approach is to focus on 4 domains (suburbs, dense urban, freeways and weather). Waymo believes that when their autonomous driving can safely handle those 4 domains that it will be good enough to scale. And Waymo has strategically chosen Chandler, SF, downtown Phoenix and now LA because they believe those areas will help them get closer to being able to safely drive in those 4 domains. Waymo also tested in NYC and Bellevue, WA to help train the Waymo Driver in the "dense urban" domain and the "weather" domain, respectively.

 

[ThomasD]

What happens to the cars when they are not picking up passengers? Do they just drive around waiting for a Fare or do they park someplace until they get a Fare? You are going to need a lot more cars during rush hour than non rush hour. If you have 5,000 cars to service an area but then you need an extra 3,000 cars at a certain time of day what happens to those extra cars when not needed? Do the cars just roam around? Returning to Depot may not be an option if the cars are leaving to city to the Suburbs farther away.

 

[diplomat33]

What happens to the cars when they are not picking up passengers? Do they just drive around waiting for a Fare or do they park someplace until they get a Fare? You are going to need a lot more cars during rush hour than non rush hour. If you have 5,000 cars to service an area but then you need an extra 3,000 cars at a certain time of day what happens to those extra cars when not needed? Do the cars just roam around? Returning to Depot may not be an option if the cars are leaving to city to the Suburbs farther away.

They go pick up the next person waiting. And yes, companies like waymo will plan for rush hour by sending out more cars from the depot. When the cars are not needed, they will return to the depot.

 

[Bladerskb]

Eastern suburb of Seattle - with a high concentration of Google, Amzn & MSFT employees

ps : Chances of me moving to a different place before Waymo (if ever) gets here is high.

Seattle is Top 2 Uber market behind SF. It’s almost guaranteed that it will be on top of Waymo’s list.

Heck Waymo is already doing small scale testing there:

https://twitter.com/x/status/1568288678975533057

How do you know ? It may fold before that ...

ps : From 2019 ...

Waymo does work in your city. You just don’t have access to it.

But Of-course because it’s much easier to develop an ADAS system that relies on the human driver and release it than a fully autonomous that has no human driver. The key is, what does the current trajectory look like and what current situations does Waymo has to solve in order to be ready to deploy in Seattle?

Remember there are 4 difficulty categories and for each category there are 3 tiers.

Categories

1) suburb
2) city
3) urban
4) highway

Tiers

1) Sunny
2) Rainy
3) Snow

Waymo has already solved suburb environment by deploying to East Valley Phoenix and City environment by deploying in Downtown Phoenix and are 70% there with solving Urban environments by deploying in SF and finally around 6 months out from offering driverless on Highways in phoenix which they would then solve.

The question to ask is, how different is your eastern suburb in Seattle compared to Phoenix. Are there some completely new topography, rules, etc that could complicate things? Not that I know of. Maybe you do. I don't know.

What I do know is that Seattle consists of Moderate to Heavy Rain and Light to Moderate Snow.
Which means that Waymo would have to conquer moderate rain and light snow in order to deploy to Seattle areas.

Your question should then be, when will Waymo go driverless anywhere in moderate rain / light snow.
And if you are saying 2040s for eastern suburb, are you saying it will take Waymo around 20 years to conquer rain/snow?

 

[Bladerskb]

But they need to be able to connect the city to suburbs using very congested freeways. BTW, have they started working on freeways ?

I think we are about 6 months (around mid 2023) till Waymo starts offering driverless on Phoenix freeways.

https://twitter.com/x/status/1590779438782124035

BTW - I don't think it is actually very difficult for Waymo to cover a city. They need to focus on the city, map it, test thoroughly and then be able to cover it. Just that there are so many cities ... that is where scaling comes. How fast can you scale ....

Now you are asking the right questions!

Waymo currently operates in about 70% of SF and based on Waymo's previous patterns. They usually pick roughly 50 sq mile of an area/city and go driverless in it.

For example they are operating in over 50sq mile in Easy Valley Phoenix. They are also operating in around ~44sq mile in Downtown Phoenix. Finally they are also operating in around 35-40sq mile in SF out of a total of 50 sq mile that SF consists of.

Here is the map for PHX

Here is the map for SF

So its save to assume that they will pick ~50sq of LA and go driverless in it.

Waymo has been spotted in LA in two recent locations:

• Washington Blvd by Marina Del Rey in Santa Monica, CA
• Colorado Ave & Ocean Ave, Santa Monica, CA

So that kinda tells us that they will attempt to pick an area in/around Santa Monica. So if we were to plot a 48 sq mile area around Santa Monica, we end up with something like this.

https://imgur.com/BMsZDvu

33.96592693739399,-118.45595800829274 34.011473386282304,-118.3154531712566 34.06176136129721,-118.35433959960938 34.09293220811057,-118.37443489172053 34.04650638443257,-118.44694280274831 34.02528303849776,-118.5147534295884

So to rephrase your question, I would ask how long do you think it would take Waymo to go driverless in that ~50 sq mile area in/around Sonta Monica, LA?

Remember it took them 13 months to go from full scale testing in SF to driverless deployment. If they cut down their full scale testing to driverless deployment time by ~50% (that would 13 months to 6 months).

So lets also assume that the amount of engineers required to make this happen were also reduced from 75% to 50%. The next question would be what happens after this deployment? The next 50sq mile area they select. How long will it take them to start full scale testing and deploy? For example Miami, Dallas or Houston. If we assume they cut that time by around ~50% again, that means they can go from full scale testing to driverless deployment in 3 months. Lets also assume they reduce the amount of engineers they require to get this done down to 25%.

That would mean by the end of 2023, they are at 3 months and 25% of engineers required to launch.

Well using those statistics, they don't have to launch in one city at a time? Their main goal up to the ~end of 2023 has been to create a driverless deployment process, master the process, reduce the resources and time for that process and then perform that process simultaneous at the same time in other locations.

With the time horizon being 3 months and 25% engineers required, they could attempt to launch in 4 cities every 3 months, which would put them at launching in 16 cities per year.

If we are alil more optimistic and say they reduce the resources (Engineers) needed to launch in a city even more, down to 10% (Plausible because more and more systems would be automated). Lets say this happens by end of 2024. They could theoretically attempt to launch in 10 cities every 3 months in 2025. That's 40 cities per year. Getting to 100 cities easily by 2028.

At that rate, logistics and infrastructure becomes the bottleneck. With this math, do you still believe your 2040s statement before they service Eastern Suburb of Seattle?

 

[diplomat33]

Waymo has already solved suburb environment by deploying to East Valley Phoenix and City environment by deploying in Downtown Phoenix and are 70% there with solving Urban environments by deploying in SF and finally around 6 months out from offering driverless on Highways in phoenix which they would then solve.

I know it is dangerous to try to predict when FSD will be "solved". But I feel like Waymo could solve the 4 domains (dense urban, suburban, highway and weather) in 2-5 years. Some might say that is crazy but as you said they have solved suburb, they are ~70% from solving dense urban and 6 months away from solving highway. And Dolgov mentioned that Waymo Driver can handle rain and wet roads so Waymo is making progress towards solving the weather domain. It just feels like Waymo is getting really close.

Additionally, I feel like Waymo is making great progress with ML from NN that can predict behavior of hundreds of objects and entire scenes, transformers, NN that can fuse radar and camera data, unsupervised learning that can detect unlabeled moving objects, new lidar perception NN that does not depend on range. It's incredible to see what AI is capable of now. The fact is that we have tools to solve autonomous driving challenges that we did not have before. Not to mention how increasingly powerful computer chips are becoming. That makes me optimistic about the odds of solving autonomous driving.

 

[EVNow]

Waymo does work in your city. You just don’t have access to it.

Show me .... (no I don't live in Bellevue .... and strangely that video doesn't seem to show Bellevue, but near Kirkland where Google office is).

Waymo currently operates in about 70% of SF and based on Waymo's previous patterns.

They had earlier specifically chosen areas of SF that didn't have dense traffic ... infact exact opposite of dense Uber routes. Has that changed ?

Remember it took them 13 months to go from full scale testing in SF to driverless deployment. If they cut down their full scale testing to driverless deployment time by ~50% (that would 13 months to 6 months).

So lets also assume that the amount of engineers required to make this happen were also reduced from 75% to 50%. The next question would be what happens after this deployment? The next 50sq mile area they select. How long will it take them to start full scale testing and deploy? For example Miami, Dallas or Houston. If we assume they cut that time by around ~50% again, that means they can go from full scale testing to driverless deployment in 3 months. Lets also assume they reduce the amount of engineers they require to get this done down to 25%.

I remember someone else thinks FSD improves exponentially

 

[Doggydogworld]

No, I don't expect people to share AV personal cars. But I do think people will share AVs like the Origin for the same reason people share buses: they are designed for many people to share.

People used (shared) buses because they're cheaper than driving and (especially) parking their own car. They sacrifice convenience and even safety to save money.

You are talking about AV personal cars.

No, I'm talking about Robotaxis.

 

[diplomat33]

People used (shared) buses because they're cheaper than driving and (especially) parking their own car. They sacrifice convenience and even safety to save money.

Yes and I expect people will use the Origin, Zoox "square pod" or Waymo-Geely driverless vehicle for the same reasons, with the added benefit of safety too.

No, I'm talking about Robotaxis.

Ok. You wrote, "They also eliminate the "time cost" of driving, since you are free to work, nap, watch porn, etc.". To be fair, that statement could apply to AV personal cars too since an AV personal car would also allow you to do other things while driving. That is why I thought you were talking about personal cars. I do agree that people will use AVs more since they don't have that "time cost". So I do expect once we have AV personal cars that people will use them more. Robotaxis cost money to use for each trip so I am not sure people will use them more. I guess it would depend on whether they are cheaper than the alternative.

 

[Doggydogworld]

Robotaxis cost money to use for each trip ....

If Robotaxis ever actually become cheap the vast majority of users will have subscriptions. There will be tiers, like the early cell phone plans. "500 weekday miles plus unlimited nights and weekends for only $299/month!"

 

[diplomat33]

If Robotaxis ever actually become cheap the vast majority of users will have subscriptions. There will be tiers, like the early cell phone plans. "500 weekday miles plus unlimited nights and weekends for only $299/month!"

I agree. I think a subscription model would work great for robotaxis. And yes a subscription model could encourage people to use robotaxis more. After all, if you are paying $299/month for a 500 weekday miles and unlimited weekends, you would want to get your money's worth.

 

[Bladerskb]

Show me .... (no I don't live in Bellevue .... and strangely that video doesn't seem to show Bellevue, but near Kirkland where Google office is).

You answered none of my questions, by attempting to answer my questions we could get to the bottom of what would actually happen in Seattle.

They had earlier specifically chosen areas of SF that didn't have dense traffic ... infact exact opposite of dense Uber routes. Has that changed ?

Hence why i said 70%. They cover 35-40sq mile out of 50 sq

I remember someone else thinks FSD improves exponentially

This comment is meanless. You answered none of my questions. And by not answering any of the questions, it results in the conversation having no foundation and going no where. I noticed this trend in alot of Tesla Fans. It helps them avoid accountability. You ask them what is 21 + 14 divided by 7 and they tell you something eqvalent to: 'didn't the weather channel say its gonna rain tomorrow?'

You gave a 2040 timeline and I asked you direct questions and scenarios regarding it. Which you refused to address.

1. The question to ask is, how different is your eastern suburb in Seattle compared to Phoenix. Are there some completely new topography, rules, etc that could complicate things?
2. When will Waymo go driverless anywhere in moderate rain / light snow?
3. If you are saying 2040s for eastern suburb, are you saying it will take Waymo around 20 years to conquer rain/snow?
4. How long do you think it would take Waymo to go driverless in that ~50 sq mile area in/around Sonta Monica, LA?
5. The next 50sq mile area (after Santa Monica, LA) they select. How long will it take them to start full scale testing and deploy?
6. Using your own math (the answers to the above quesitons), can they simultaneously launch in more than one city at a time?
7. Using your own math do you still believe your 2040s statement before they service Eastern Suburb of Seattle?

 

[Krispykreme]

You answered none of my questions, by attempting to answer my questions we could get to the bottom of what would actually happen in Seattle.

Hence why i said 70%. They cover 35-40sq mile out of 50 sq

This comment is meanless. You answered none of my questions. And by not answering any of the questions, it results in the conversation having no foundation and going no where. I noticed this trend in alot of Tesla Fans. It helps them avoid accountability. You ask them what is 21 + 14 divided by 7 and they tell you something eqvalent to: 'didn't the weather channel say its gonna rain tomorrow?'

You gave a 2040 timeline and I asked you direct questions and scenarios regarding it. Which you refused to address.

1. The question to ask is, how different is your eastern suburb in Seattle compared to Phoenix. Are there some completely new topography, rules, etc that could complicate things?
2. When will Waymo go driverless anywhere in moderate rain / light snow?
3. If you are saying 2040s for eastern suburb, are you saying it will take Waymo around 20 years to conquer rain/snow?
4. How long do you think it would take Waymo to go driverless in that ~50 sq mile area in/around Sonta Monica, LA?
5. The next 50sq mile area (after Santa Monica, LA) they select. How long will it take them to start full scale testing and deploy?
6. Using your own math (the answers to the above quesitons), can they simultaneously launch in more than one city at a time?
7. Using your own math do you still believe your 2040s statement before they service Eastern Suburb of Seattle?

I think tesla fan totally misunderstood how certification works for self driving permit.

1. In US because there is no federal laws in self driving. All of their permitting and granting of operating license is done at city/state level.

2. For Waymo to get into particular city- it has to agree with that city on where to do the sand box testing. Only when Waymo/cruise/Zoox demonstrate reliability in operating within that sand box. Then commercial operation permit will be awarded.

3. City chose the sandbox area not Waymo or cruise .

Waymo will launch more and more faster because its system is more robust and ML is actually way better trained from day one because it was trained on L4 data set with multiple object detection/ classification system. Hence you see LA launching soon not long after SF.