Connected/Automated vehicles and infrastructure in Boston

Stlin said:
I use waymos in preference to human driven ubers when I'm in SF - for the most part, they're pretty good about the limitations of curb access - youre requesting a ride from a spot we can't stop - go here to get your waymo. It actually usually wouldn't let me select my location, and it would usally give me a selection of nearest pickup points where it judges it can safely stop in downtownier bits of SF. Same thing with dropoffs - it consistently will pick the nearest intersection.

Theres a few situations where they struggle - they're risk adverse, and in certain situations can't make the decisions a human driver would in a crowded pickup situation. I will say I never felt unsafe as a human around them, nor as a passenger in them.
Click to expand...
This sounds like an ADA nightmare
 
KCasiglio said:
This sounds like an ADA nightmare
Click to expand...
Ngl, not something that had occurred to me, so I just checked the app pretending I need an accessible ride from Moscone. So apparently thats a feature you can turn on - theres a somewhat hidden toggle in the accessibility settings for "minimize walking time: When possible, shorten walk at pickup and dropoff, including preventing the need to cross the street. A shorter walk may mean a longer ride overall."

Also, apparently there's an option for a Human driven waymo via requesting a wheelchair accessible vehicle - i guess for the need to deploy the ramp & secure the wheelchair.
 
KCasiglio said:
people seem to like them in SF and Austin from the little I've looked into it.
Click to expand...
A lot of people like the novelty - but - a lot of people in SF hate them because they crash out in the most inopportune times and locations. This kind of video pops up in my group chat with some friends who live in the Bay Area about once or twice a month and people mock the cars.

There was a congressional hearing the other day as well and it sounds like much of Waymo's operations are outsourced to the Phillipines. No problem with Filipinos, but, the largest concern was that Waymo was a bit evasive on how often human intervention is required for regular operations and whether the interveners are licenses or aware of all the situations and how locals navigate the conflicts.

The hearing is also in part because of a recent California crash during school hours where the Waymo "Driver" ran into a child that entered the roadway. One of the Waymo "remote" drivers also ran into parked cars in California recently as well. Both have been pretty big stories in cities where Waymo has supportive, fair weather conditions.
 
as02155 said:
A lot of people like the novelty - but - a lot of people in SF hate them because they crash out in the most inopportune times and locations. This kind of video pops up in my group chat with some friends who live in the Bay Area about once or twice a month and people mock the cars.

There was a congressional hearing the other day as well and it sounds like much of Waymo's operations are outsourced to the Phillipines. No problem with Filipinos, but, the largest concern was that Waymo was a bit evasive on how often human intervention is required for regular operations and whether the interveners are licenses or aware of all the situations and how locals navigate the conflicts.

The hearing is also in part because of a recent California crash during school hours where the Waymo "Driver" ran into a child that entered the roadway. One of the Waymo "remote" drivers also ran into parked cars in California recently as well. Both have been pretty big stories in cities where Waymo has supportive, fair weather conditions.
Click to expand...
I mean I'm not terribly bothered by the human intervention aspect. Despite the flaws one human can still 'drive' multiple cars which has obviously not been possible before.
 
TheRatmeister said:
human intervention aspect.
Click to expand...
Given that they're already starting from a place of outsourcing the monitor and intervening process - I do wonder how much, as Waymo scales, especially to places like Boston, this may mean financially sustainablity is dependent on a Matrix-like farm of cubicle workers monitoring and intervening.
 
As a current SF resident, the Waymos are pretty awful. They almost entirely pick up and drop off in travel lanes - there does not seem to be any effort to have them pull to the curb even if space is available. Any pickup/dropoff on a busy street blocks traffic and transit. That also means they don't lay over between trips; they simply keep driving. (Anecdotally, less than a quarter of Waymos I see have anyone in them).

Minor glitches, where the vehicles stall for a few seconds, seem frequent. The company refuses to provide any data about even the biggest incidents like the behavior during the power outage.
 
The EGE said:
Minor glitches
Click to expand...
I haven't heard about this aspect from others. Is this like when people wait too long to start after getting the green light or does it last a bit longer? Or is it like when my friend told me they got unexpected trapped in one - and they had to "SOS" to get a remote operator to open the door since the car wasn't going anywhere but also not unlocking doors either.
 
It seems to be random, often midblock, as if the computer locked up and a human took over. No obvious pattern from what I've seen.
 
The EGE said:
As a current SF resident, the Waymos are pretty awful. They almost entirely pick up and drop off in travel lanes - there does not seem to be any effort to have them pull to the curb even if space is available. Any pickup/dropoff on a busy street blocks traffic and transit. That also means they don't lay over between trips; they simply keep driving. (Anecdotally, less than a quarter of Waymos I see have anyone in them).

Minor glitches, where the vehicles stall for a few seconds, seem frequent. The company refuses to provide any data about even the biggest incidents like the behavior during the power outage.
Click to expand...
Not an SF resident, so take it with a grain of salt, but last time I visited I rode the city bikes all over and loved the Waymos. They acted predictably, drove slowly, stayed out of bike lanes, gave me plenty of space when passing, and never rolled down their windows and screamed threats at when I used a full travel lane.
 
ball_square said:
They acted predictably, drove slowly, stayed out of bike lanes, gave me plenty of space when passing, and never rolled down their windows and screamed threats at when I used a full travel lane.
Click to expand...
This is the vast majority of drivers, as well.
 
as02155 said:
This is the vast majority of drivers, as well.
Click to expand...
Sure, most human drivers, but all Waymo drivers.

The difference between "95% chance you won't endanger my life" and "100% chance you won't endanger my life" is stark when you interact with hundreds of cars per ride.

(anyway, 50%+ of human drivers seem to go over the speed limit when there is no traffic calming)
 
ball_square said:
Sure, most human drivers, but all Waymo drivers.

The difference between "95% chance you won't endanger my life" and "100% chance you won't endanger my life" is stark when you interact with hundreds of cars per ride.

(anyway, 50%+ of human drivers seem to go over the speed limit when there is no traffic calming)
Click to expand...
Sure, but, based on crash statistics - it's like 96% vs 98%. Waymo still doesn't seem to recognize children very well. And Waymo is seemingly the best.
 
as02155 said:
Given that they're already starting from a place of outsourcing the monitor and intervening process - I do wonder how much, as Waymo scales, especially to places like Boston, this may mean financially sustainablity is dependent on a Matrix-like farm of cubicle workers monitoring and intervening.
Click to expand...
For what it's worth, it looks like they've got 70 employees that do the remote help on 2-3000 vehicles. The technology is obviously in it's infancy, as they gain more real world use information or improve the underlying algorithms, I bet the supervisor to taxi ratio should improve hugely.

 
as02155 said:
What indication do you have that this will be the case? Algorithmic improvements seem to have a decay interval that often lead to bankrupcy post VC series 3.
Click to expand...
I'm no automation engineer but they started from needing one to one in person supervision but are now down to only needing one person to manage 40 cars at once over the course of the last few years. All of that while now achieving crash rates substantially below human level. I think the presumption should be that performance will continue to improve with the deployment set to expand hugely in 2026. If the rationale for needing occasional human intervention is the very long tail of rare traffic scenarios, then as we go from the current 200 million autonomous miles to 200 billion, the orders of magnitude more data for future training runs should gradually move scenarios from the human to autonomous pile.

But putting that aside, Uber could scale with 1:1 supervision paid at (high in global terms) local wages at all times for each mile driven. Even if there were no future improvements, which I find implausible, 1:40 supervision paid at low-medium global wages for each mile seems like a bargain in comparison regarding your earlier concern about the sustainability of specifically the cubicle operation.
 
You must log in or register to reply here.

Back
Top