back
24 comments
Part of the reason AVs are not a panacea. Walkable cities are fault-tolerant cities. No way around that.
Indeed. The picture in the article just shows how inefficient cars are in cities. A bicycle could easily filter through that entire line of cars in under a minute.
You're not wrong but 19th and Ulloa is the city portion of highway that connects to the golden gate bridge so many of these drivers are possibly just trying to pass through the city from outside

edit spelling

Which do you think will happen first, SF has 100+ driverless AVs running simultaneously or a new protected bike lane is built in the inner sunset?
this comes a few days after Waymo went through some pretty big layoffs, I'm not sure if the "rider assist team" mentioned in the article was part of the cut, if they were I'd hate to be a part of the team working that closely after a team wipe like that
Fascinating how slowly driverless cars have progressed in the past 5-10 years.
It really shows who the sober voices were: the people selling huge valuations had billion-dollar marketing marketing and were able to drown them out, but it’s not like there weren’t plenty of skeptics.

Unfortunately, there’s pare some powerful politics behind the dream: cars are inherently unscalable but everyone with an expensive suburban house wants to believe otherwise and since transit is coded poor in the U.S. politicians really liked the idea that they could do what the politically-powerful wanted by doing nothing they weren’t already doing. Why put effort into making buses efficient and annoying a bunch of car commuters when you can say FSD will fix it?

As a friend of mine who's heavily involved in transit pointed out, we pretty much nailed autonomous self-driving subways a long time ago. But as you say, in the USA, mass transit is for poors.
They should have remote operators on call. It should be simple to integrate a routing into the software that sends an alarm if it detects the car is stopped in the middle of the road for more than 5 minutes. Remote operator gets the alarm, takes control and moves over the car to the side of the road.
The self report of the deputy managing editor (link in the article) is IMO ways more interesting. The question “for what people serves this kind of technology” is a non-brainer, though. The traffic (and so the world) would be much better and safer if all vehicles are self-driving (and inter-connected).
Been meaning to get a set of these:

https://www.harborfreight.com/1250-lb-capacity-vehicle-posit...

Seriously, this is not a complex intersection and Google has no business beta testing on public roads.

I watched those used in real-time as a jerk left his car at the entrance to a casino. A valet ran to get those and they just moved his car out of the way. Restored my faith in humanity how smooth of an operation it all was.
Would those castors roll on an average road? Looks like they're intended for a smooth concrete floor. If they do roll easily and the road is not perfectly level, the car might even roll away. An even more dangerous form of driverless car!
From the review comments:

  “The only con I can name is the casters do not roll as smoothly as better quality casters. In moving a ‘63 Galaxie 500 it takes 2 people to make it roll on a perfectly smooth and swept floor.”
I can't be bothered to actually make the meme right now, but imagine I used this for "most complex intersection in SF" and put it next to "simplest intersection in Boston".
Look up St. Francis Circle which is less than a mile away.
Don't need to, been there in person. My comment wasn't meant to be taken literally.
I agree. They can develop and test the perception system on public roads using a vehicle with a human driver, and the controller on private roads.
From a computation standpoint dealing with "unexpected road construction" seems like a tough issue but from a practical standpoint it's something drivers in many cities have to deal with basically every single day. As least the vehicle just stopped as opposed to causing any other issues
I wonder how self driving cars would manage a "letting an emergency vehicle through"-situation, especially during a traffic jam or at a red light. Because human drivers do the right thing there and do "forbidden" stuff there, like ignoring lanes or driving onto side walks and such.
That's a good point and one i'm curious about too. Just today I was on a normal route driving and came across a road closure behind a blind turn because of an accident - i'm not a fan of self driving personally
How would anyone know? Traffic is always at a halt in San Francisco.
If the road is empty before that thing and full after it, that might be a sign.
It might be enlightening if you, as I just did, look up that intersection on your mapping service of choice. "Very complex" it is not.