Miles driven is only one axis of many, does Waymo get from A to B as quickly as the average human driver would, or does it play it safe by staying in the slow lane and taking suboptimal routes to avoid challenging areas?
Another bias is baked into this announcement: how would it fare in Boston, NYC, or DC? Are they better because they're better than humans or are they better because they're cherry-picking their test cities?
San Francisco is not Manhattan but its hardly an ideal driving environment. It’s a lot harder to drive there than Phoenix.
Narrow streets packed with park cars, lots of bikes and pedestrians jaywalking, a homeless population that spills over into infrastructure, etc. It’s a good test case for most US cities that aren’t NYC.
And lets not forget the hills. Driving in SF is no joke. In my 20+ rides with Waymo, it feels that it drives more confidently than most Uber drivers I've had experience with.
In SF, waymo isn't particularly slow, and drives remarkably humanlike to me, unlike Cruise which regularly took 2x as long to drive anywhere as a human would, and drove very much not like a human would.
3.5x better at injuries, 10x the capex per vehicle, 100x the operational capex of a normal taxi service. Plus you get to own the vehicles, pay costs for all maintenance, store them when not in use, carry all liability.
We need actual mass transit that works for people, not a roving science experiment on wheels.
I see buses driving around my city all the time. They're noisy, pollute, take up space, and often have one or two people on them.
Why do you think the TCO is less for a bus than for a self driving car? Why would it be more efficient to have mass transit when you don't have extremely high density? Can you share some data?
He might have meant trains rather than buses when he brought up mass transit. It's rare I've been on a train in a metro area only carrying one or two people.
If self driving cars were really capital efficient, why hasn't Google replaced their fleet of busses that travel the south bay to SF with their self driving cars? It's not because of anything technical, it's because it doesn't make any financial sense to replace a bus (even one that carries a few people) with a self-driving car.
That, and the business unit that actually operates the shuttle busses lives in the real world.
Google's buses looked pretty full a lot of the time (a few years ago at least). The economics of those are probably pretty good. That said: I doubt city buses are that bad either. Data is needed for both of these points, speculation is a waste of time.
the business unit that actually operates the waymos operate in the real world, and so are subject to real life scaling problems. Assume the self driving cars are capital efficient and could actually make the drive from San Francisco to Mountain view (they legally are allowed to, but Waymo has not unlocked this ability to the general public). yYu still need to manufacture the damn things, and Google isn't a car company. Manufacturing is hard. Tesla making 5,000 Model 3's that fist quarter was considered an impossible accomplishment. Google is reliant on many third parties to produce Waymos. How long do you think is should take them to replace busses assuming those priors?
It's gonna take some time. Regardless of the economics of it, they can't just magic up a fleet of Waymos. That production has been planned for and allocated and remember, they don't actually offer rides down the peninsula that the shuttle busses do yet. That said, I can't wait for Waymo technology to mature and be cheap enough for cities to use them on mini buses.
Maybe your city should invest in better buses, road infrastructure (eg bus lanes) so they not stuck in traffic, and optimize the schedule and then you will see that the public transit is actually a more efficient alternative to self driving taxis
Can you clarify your argument? Is it that "no, buses are not more efficient, but with enough investment they can be?".
Do you have some data to share to show this is true in suburban environments? I live in a very wealthy area, budget is not a problem for the city, the busses are clean and well run. But they're often empty...
> road infrastructure (eg bus lanes) so they not stuck in traffic
Bus lanes aren't going to do anything to help with the real bottleneck in city bus transport: having to stop every couple of blocks to let people on and off.
I don’t know whether the specifics of your numbers are true or not, but if efficient transit for the masses in urban areas is the goal, self-driving cars have proven they are not the solution.
Even with a 100% safety rating, cars don’t scale to denser urban environments such as SF, NYC, Boston, Seattle, etc… For that we need buses, trams, and subways. My preference would be self-driving dedicated bus rapid transit lanes.
No you don’t. AVs with 2-4 riders provide excellent passenger density and provide way higher quality service than traditional mass transit. Trip times are typically less than half of even a good mass transit system, and it’s a far more relaxed environment.
The biggest problem with mass transit is that it is slow, and while you can improve a system to make it faster, there are ultimate necessary limits because the passenger has to walk to a stop, wait for a vehicle, and then wait while the vehicle stops and passengers get on and off, and then walk to their destination. You cannot fix this, it’s inherent to the medium.
I seem to recall a startup or two working on AV ‘micro-buses’. That is, AV buses that held 4-8 passengers and that would have broader route patterns than traditional buses. Anything ever happen to those? I can see how those could complement traditional mass transit very well.
There are a few. Zoox comes to mind in that their vehicle is somewhat bus-like in that it has bus type doors, and somewhat bus like seating and interior (plastic/vinyl, easy to clean), but it only holds 4 passengers.
Uber and Lyft already have a similar (with a driver) service that they've rejiggered a few times. I used it a fair bit pre-covid to commute. The price wasn't much more than driving myself, and the speed was good -- usually it was only about 5 minutes slower on a 30-40 minute drive, and I could use the time to catch up on email, read, etc. The same ride on transit would have been about 1.5 hours, or 1 hour if I took a bike on the train instead of using busses or walking for part of the trip. Transit was also a bunch of broken up blocks of time either standing waiting, or sitting on a train or getting on and off/in/out of a station. With the shared ride vehicle, it was just wait a few minutes for pickup and then I had a solid block of time in a fairly comfortable environment.
I really think that this is the future of transit. If you have enough usage, you could probably push these vehicles up to 8-10 people and still get reasonable travel times (way better than transit) at transit-competitive costs. People who wanted better speed and more privacy could order smaller vehicles for a higher price. If these were popular it would massively reduce congestion because people would actually want to take them instead of driving themselves.
In the US, with its suburbs and spread out cities, AV cars, microbuses and the lot definitely have a role to play. There’s no way around that.
Where was your experience for mass transit? The US? I live in Boston, which has one of the best mass transit systems in the US, but still it sucks as a general means to get around, or even to and from work for many people. But the US has extraordinarily poor mass transit overall. Budapest, Paris, Tokyo, Vienna, etc… have efficient, multi-layered, redundant mass transit that serves the population well as a general means to get around. It’s faster than the auto frequently even in non rush hour times.
I don’t see (good) mass transit and AVs at odds with one another, but rather complimentary. Every auto off the road makes the remaining auto transit more efficient. Less autos on the road also frees up roads for bikes and pedestrian use, a big plus in denser urban environments in need of more green and public space.
This was in the Bay Area, which has OK but not great transit. I used to live in Berlin though that has great transit, which I rode a lot, and it's still much slower than driving or taking a Taxi.
Ultimately you just cannot get around the physics of it. If you have a 50 person vehicle, the average person is going to need to wait for 50 people to get on or off during their ride. In order to do that, they're going to have to make a bunch of stops. To do that, the vehicle has to slow down and speed up at a reasonable acceleration rate, so the average speed is necessarily much slower than the maximum speed.
With fixed routes, you also cannot get around having to walk to and from the station/stop, and you can't get around having to wait at least some time for the vehicle to arrive. For most trips you also can't avoid having to do a transfer with more walking/waiting.
The solution is less passengers per vehicle so there's less stopping, and dynamic routes so that there's less walking and less waiting. You can't really do door to door dynamic routes with lots of passengers, because you'd be going too far out of your way and it would be slow.
This wasn't possible before because you had to pay a driver, but if a driver is low cost, the game is completely different.
I love it. It’s clean. Quiet. Relaxing. And I don’t have to deal with other people. You couldn’t pay me to take public transportation over a Waymo. What a joke.
I can’t wait until this tech makes it to private vehicles.
If the technology matures and is mass produced, those costs are going to come way down.
You probably don't like single/low occupancy vehicles, and that's fine, I'm not here to convince you otherwise. But this particular argument is nonsense, it applies to every new technology ever.
If your human bus driver crashes you place all the blame on that guy, fire him and move on with your business. If your AI that is used to drive every vehicle your fleet screws up you're potentially looking at grounding the entire fleet and running an extended PR campaign.
At the very least you'll need to stick a human behind the wheel to take the blame when the technology fails like Tesla does. You can't remove humans from the corporation centered blame shifting society because you remove available scapegoats leaving the company without a blame shield. Perhaps they'll hire third world people for lower wages to remotely watch the AI car feed all day and auto-fire them in the event of a crash.
This is all still a worse deal than Uber hiring humans to do it all, they outsource all of the maintenance and liability onto some rando that they can ban from the app if they get uppity and then take a cut of the profit in exchange for running a data center and some ads.
Something I haven't considered much but, in a world where you hop into a self-driving taxi; how is the passenger protected from someone who wants to rob them?
I'm imaging a situation where an attacker can step in front of the vehicle, have it safely stop, before demanding you hand over your valuables, breaking a window to get in etc.
As a passenger you're trapped in an enclosed space and left without a way to escape, where as a human driver would be able to make the judgement call to do whatever is necessary to get out of danger.
Yeah but why would a passenger in a locked car that's loaded with cameras and connected to a call center be considered an easy target? And it's hard to tell if there's even someone in the Waymo because of the tinted windows.
I encountered this problem in a Cruise. I was picked up late at night with a friend, and this drunk person ran over and banged on the window, which caused the Cruise to panic and stop. It called customer support and I told them that there were some drunk people and to please get us out of there. The car started moving, thankfully, but it's an interesting problem to consider.
38 comments
[ 2.9 ms ] story [ 85.1 ms ] threadNarrow streets packed with park cars, lots of bikes and pedestrians jaywalking, a homeless population that spills over into infrastructure, etc. It’s a good test case for most US cities that aren’t NYC.
A LIDAR is pretty much all you need to avoid a collision. It's a pretty forgiving environment.
We need actual mass transit that works for people, not a roving science experiment on wheels.
Why do you think the TCO is less for a bus than for a self driving car? Why would it be more efficient to have mass transit when you don't have extremely high density? Can you share some data?
That, and the business unit that actually operates the shuttle busses lives in the real world.
They ARE slowly replacing busses with ride share and will probably move to self driving once the city approves it.
It's gonna take some time. Regardless of the economics of it, they can't just magic up a fleet of Waymos. That production has been planned for and allocated and remember, they don't actually offer rides down the peninsula that the shuttle busses do yet. That said, I can't wait for Waymo technology to mature and be cheap enough for cities to use them on mini buses.
Do you have some data to share to show this is true in suburban environments? I live in a very wealthy area, budget is not a problem for the city, the busses are clean and well run. But they're often empty...
Bus lanes aren't going to do anything to help with the real bottleneck in city bus transport: having to stop every couple of blocks to let people on and off.
Even with a 100% safety rating, cars don’t scale to denser urban environments such as SF, NYC, Boston, Seattle, etc… For that we need buses, trams, and subways. My preference would be self-driving dedicated bus rapid transit lanes.
The biggest problem with mass transit is that it is slow, and while you can improve a system to make it faster, there are ultimate necessary limits because the passenger has to walk to a stop, wait for a vehicle, and then wait while the vehicle stops and passengers get on and off, and then walk to their destination. You cannot fix this, it’s inherent to the medium.
Uber and Lyft already have a similar (with a driver) service that they've rejiggered a few times. I used it a fair bit pre-covid to commute. The price wasn't much more than driving myself, and the speed was good -- usually it was only about 5 minutes slower on a 30-40 minute drive, and I could use the time to catch up on email, read, etc. The same ride on transit would have been about 1.5 hours, or 1 hour if I took a bike on the train instead of using busses or walking for part of the trip. Transit was also a bunch of broken up blocks of time either standing waiting, or sitting on a train or getting on and off/in/out of a station. With the shared ride vehicle, it was just wait a few minutes for pickup and then I had a solid block of time in a fairly comfortable environment.
I really think that this is the future of transit. If you have enough usage, you could probably push these vehicles up to 8-10 people and still get reasonable travel times (way better than transit) at transit-competitive costs. People who wanted better speed and more privacy could order smaller vehicles for a higher price. If these were popular it would massively reduce congestion because people would actually want to take them instead of driving themselves.
Where was your experience for mass transit? The US? I live in Boston, which has one of the best mass transit systems in the US, but still it sucks as a general means to get around, or even to and from work for many people. But the US has extraordinarily poor mass transit overall. Budapest, Paris, Tokyo, Vienna, etc… have efficient, multi-layered, redundant mass transit that serves the population well as a general means to get around. It’s faster than the auto frequently even in non rush hour times.
I don’t see (good) mass transit and AVs at odds with one another, but rather complimentary. Every auto off the road makes the remaining auto transit more efficient. Less autos on the road also frees up roads for bikes and pedestrian use, a big plus in denser urban environments in need of more green and public space.
Ultimately you just cannot get around the physics of it. If you have a 50 person vehicle, the average person is going to need to wait for 50 people to get on or off during their ride. In order to do that, they're going to have to make a bunch of stops. To do that, the vehicle has to slow down and speed up at a reasonable acceleration rate, so the average speed is necessarily much slower than the maximum speed.
With fixed routes, you also cannot get around having to walk to and from the station/stop, and you can't get around having to wait at least some time for the vehicle to arrive. For most trips you also can't avoid having to do a transfer with more walking/waiting.
The solution is less passengers per vehicle so there's less stopping, and dynamic routes so that there's less walking and less waiting. You can't really do door to door dynamic routes with lots of passengers, because you'd be going too far out of your way and it would be slow.
This wasn't possible before because you had to pay a driver, but if a driver is low cost, the game is completely different.
I can’t wait until this tech makes it to private vehicles.
You probably don't like single/low occupancy vehicles, and that's fine, I'm not here to convince you otherwise. But this particular argument is nonsense, it applies to every new technology ever.
At the very least you'll need to stick a human behind the wheel to take the blame when the technology fails like Tesla does. You can't remove humans from the corporation centered blame shifting society because you remove available scapegoats leaving the company without a blame shield. Perhaps they'll hire third world people for lower wages to remotely watch the AI car feed all day and auto-fire them in the event of a crash.
This is all still a worse deal than Uber hiring humans to do it all, they outsource all of the maintenance and liability onto some rando that they can ban from the app if they get uppity and then take a cut of the profit in exchange for running a data center and some ads.
I'm imaging a situation where an attacker can step in front of the vehicle, have it safely stop, before demanding you hand over your valuables, breaking a window to get in etc.
As a passenger you're trapped in an enclosed space and left without a way to escape, where as a human driver would be able to make the judgement call to do whatever is necessary to get out of danger.