back
909 comments
"Skyrocket" is a word people use when they don't have the data to support their claim, but want to make it sound as extreme as possible.

When I hear "skyrocket" I think "several orders of magnitude in a very short time". But the source of the quote with skyrocket, basically admits that they didn't have any quantitative evidence to support a massive change in incidents.

Remember: reality is banal. Things are far better explained by sampling error, bias, and base rates, than they are by sudden, dramatic shifts caused by a single factor.

Absolutely right. To situate in terms of recent events, there have been three stages of trying to delay / influence the California CPUC robotaxi expansion approval vote:

1. Tried a politician scare campaign. Failed because it was too transparent it was an attempted distraction from their terrible public safety record.

2. Tried coning and disabling the vehicles. Failed because it was too transparently astroturfed, legal liability.

3. Now attempting leaks of cherry-picked vanity metrics so massaged they can’t fairly be called “statistics” to friendly anti-tech local media.

Will fail because national and international media no longer trust our local journalists, and will investigate independently.

The cars work. The incident data meets the standards set forth by the state of California.

Local politicians need to find another anti-tech boogeyman. This one is too vital to the revitalization of SF.

I am not talking my book, have zero stake here.

It’s just saddening to watch a nascent technology finally start working well and then get kneecapped by political games.

When I hear "skyrocket" I think "several orders of magnitude in a very short time".

The article actually has quantitative data that shows reports of "incidents where driverless cars disrupt traffic, transit and emergency responders" rising by a couple orders of magnitude in a year.

Yeah, the city officials have to cover their asses and make disclaimers about what they can and cannot conclude, because all they know is that they're getting more reports of incidents; they don't have access to Waymo's data. But given that we know driverless car activity has increased substantially, it seems silly to assume that it must just be a random coincidence that people are now reporting correspondingly more problems.

I'm with ya normally, but literally a few paragraphs in they have the chart source: the TA, going from 3 to 91 "incidents" over a year. If you turn it on its side, it kinda looks like rocket blasting off.
Data appears to be freely available here: https://www.dmv.ca.gov/portal/vehicle-industry-services/auto....

I reviewed a couple dozen incidents and in every case human drivers are to blame. The most extreme example of this was an incident in which a manually operated vehicle repeatedly rear-ended an autonomous vehicle out of spite, then followed it back to the dispatch lot and made a variety of verbal threats.

The numbers they gave is more than a doubling, while they were quite clear about the limitations of their data. That includes not knowing how many AVs are on the road, which is something that the companies should be able to provide.

Also look at the claims that AVs have not caused death or serious injury. (I noticed that they did not claim that AVs have not caused injury.) That is great, except I would not expect any given corporate vehicle fleet, autonomous or human controlled, to have a record of causing death or serious injury over a short period of time. The sample size is just too small.

Either way, there's not enough data to prove anything. On the other hand, we have a group with a clear conflict of interest (the makers of AVs) up against a group which has noted concerning incidents but has not been provided with the data they need even when it should be available. Then we have a third group who are being asked to let an experiment on the general public proceed. I doubt that it would pass many academic ethics committees, but if you have money, well, go ahead!

The SF Chronicle has an unfortunate habit of headlines telling you "what to feel" as opposed to "what happened".
Waymo and Cruise won't actually release comprehensive data. So anecdotes are all we have to go by... and they don't look particularly good. These are egregious driving mistakes that a human driver would almost never make.

- https://twitter.com/desertflyer/status/1677464706251128832

- https://twitter.com/flrent/status/1677483882109882368

Everyone needs to experience these things for themselves before unleashing them on the public. I want them to work, but in their current state, they are an absolute menace.

The article includes a month-by-month graph of incident counts, broken down by company, that shows the "skyrocketing" behavior starting in March of this year.

The article notes that the data is incomplete, but that only means that there are more incidents than the graph shows.

And I'd also expect that part of the reason for the increase is that there are many more miles driven now by AVs than a year ago. But in a way that doesn't matter: an absolute increase in incidents is a problem, regardless of how much driving is going on.

This is especially the case when we're talking about things that human drivers are less likely to do, like driving through caution tape and snagging Muni wires. Not sure how often human drivers run over fire hoses or drive directly into active fire scenes, but I'd expect it's not often when compared with AV software that seems to just not know it's supposed to avoid those things.

"Self driving cars were born in San Francisco. Companies like Cruise and Waymo are creating something that will literally save lives. But some San Francisco politicians hate technology so much, they’re literally willing to make up statistics lying to the public to justify banning them. It’s a lesson in killing the golden goose." - Garry Tan https://youtu.be/rjgUPUKD-Sc
I'm wary about using public roads to test these, but I think the way the data is presented is misleading. I'm not sure how it's misleading, but separating "incidents" into categories (safety, traffic, accident, etc) might be a good start.

For example, I could start coning cruise cars, and cause these numbers to skyrocket. While that's an inconvenience to other drivers, it's not a safety issue at all.

By the way, as a motorcyclist (and thus hyper annoyed at bad driving), I find Uber/Lyft/Food drivers to be both much more dangerous and inconveniencing than these self driving cars. We should have scrutiny on the safety of these machines, but they're hardly the biggest problem facing SF.

In my experience living in SF, these cars are much safer than the typical drivers in my neighborhood for pedestrians. Most human drivers around here don't even bother to stop at stop signs, and instead just slow slightly. With the cruise and Waymo cars, I feel like the risk is a bit lower when I'm on my bike or walking.

I have witnessed a Cruise car stopping in the middle of the road when faced with an oncoming emergency vehicle, so I totally buy that they aren't ready for prime time yet.

Honestly, I'd prefer if we prioritized enforcing existing traffic laws for regular vehicles.

>Cruise and Waymo say city officials have mischaracterized their safety track records. Their driverless taxis, the companies say, have lower collision rates than human drivers and public transit.

This is comparing the mean driverless incident rate to the mean human driver incident rate. This is disingenuous, since a small minority of human drivers cause the vast majority of incidents, thereby severely inflating the mean human incident rate.

Today's self-driving cars may be safer than the mean human driver, but I would wager they are far from the median human driver, and absolutely nowhere close to the top 10% of human drivers. It may be possible (but very difficult) to beat the median human with dramatic improvements to current self-driving algorithms, but beating the top 10% of human drivers will require AGI.

I left SF a year ago but had Cruise and Waymo cars in my neighborhood constantly. They were a total nuisance. They'd jerk towards pedestrians in crosswalks leading to this uncomfortable standoff where no one in the neighborhood would ever step in front of one. I'd wager any stats here are way lower than real incidences as I was nearly hit multiple times and never got to the "report it" stage as there are so many barriers to doing so.
@garry did a rebuttal video to this. https://www.youtube.com/watch?v=rjgUPUKD-Sc
Beware: the State sent a letter accusing the city of manipulating data in order to make AVs looks more dangerous than they were: https://twitter.com/annatonger/status/1673403230804385813

> In the 4 Waymo traffic collisions SF cited to prove driverless cars are less safe, 3 were the Waymos being rear-ended and 1 didn't even involve any cars touching, CA says. https://docs.cpuc.ca.gov/PublishedDocs/Published/G000/M512/K...

Laws are written and designed for humans who constantly try to push it to the limit and get away with things. These cars are all over the place and I've been caught stuck on a street in Bernal Heights because it couldn't quite fit and just froze in place. I had to back out of the street because there wasn't any room for me to go around it.

Yes it may be following laws to the letter, but not only was I grumpy about having to carefully back out of a street, there wasn't any way to communicate with the car. It doesn't have any feedback mechanism for me, so it just looks like a confused dead robot in the middle of the street.

I think Waymo and Cruise are both happily working on hard problems in this space with limitless VC funding (or daddy GM money) thrown their way. So I support their efforts, because at the very least, they'll enrich our understanding of self-driving cars even if it were to fail in the marketplace long term.

I'm curious how profitable it will be in practice once the VC money dries up. Paying half a million dollar salaries to employees who need to keep these things up to date adds up. That, on top of hardware costs for precise sensors that will fail fast in harsh weather conditions. On top of normal wear and tear of people in and out of the cars all day for cars that won't need down time.

Lastly, I'll end my thoughts with this... why are we obsessed with the idea of self driving cars anyway?! What human problem does this solve? I am 100% on board with building autonomous cruise-control for my own car, so I'm not against the tech. I just find the taxi angle to be weird. All I see so far in my mind is a product that just takes away an entire class of job opportunities with no actual gain.

So it's funny, when they first started rolling these out folks were interested, intrigued even. Folks walking would give them deference, mostly not run in front of them if they saw them, very unlike how they treat regular drivers. But now folks are getting sick of them.

So they were modeling all their behavior and driving around cooperative pedestrians and cars (mostly) now they have to model for an adversarial public.

Good fucking luck.

Is this what Garry Tan's video (The Truth and Lies About Driverless Cars in SF [1]) was all about? I find him to be pretty credible in general, and he makes a strong case that there are some folks within the city government who are misrepresenting the safety record of self-driving cars. I don't know much about the situation, and I'm glad this testing isn't happening where I live (we do have tons of testing on the peninsula, but it's with humans in the vehicle). But I think it's important for companies and regulators to be honest about what is happening, whichever way the facts come out.

1: https://www.youtube.com/watch?v=rjgUPUKD-Sc

I lost count of how many times there was a comment about a lack of data

> The city is left in the dark about the exact number of driverless taxis operating in its streets, and the miles they’ve traveled.

> city transportation leaders this time around informally collect their own incident data using 911 and 311 calls

> Friedlander said the city can’t make definitive conclusions because it doesn’t have detailed data.

Yet Google and GM say things like

> “Every single day of delay in deploying this life-saving autonomous driving technology has critical impacts on road safety,” Waymo said in a statement.

On the face of it Google and GM are lying

These companies cannot be trusted to be honest about incidents or near incidents. They like to crow about safety, but the safety they want credit for is entirely hypothetical.

There is no good reason to believe that autonomous vehicles are actually safer. They haven’t yet been tested enough, and the testing that has been done cannot be generalized to new situations.

I believe they would be safer if all roads were fully digitized, controlled, with no pedestrians and no other human drivers. We have something like that, now, called railroads.

Mass transit is the better investment.

> In San Francisco, those problems can mean self-driving cars blocking traffic, transit and emergency responders, as well as erratic behavior resulting in close calls with cyclists, pedestrians or other vehicles.

I have mixed feelings about how to test self-driving vehicles (mostly stemming from ignorance). But at some point don't you have to get this stuff out in the wild to see how it behaves or else we are resigned to not making progress on this (or very very slow progress)? And considering that "no one was hurt" and "driverless taxis have never killed or seriously injured anyone in the millions of miles they’ve traveled" are were now there?

Can anybody find the data set from which they created the graph?

I tried browsing https://data.sfgov.org/browse but there's nothing that immediately refers to "autonomous vehicles", "incidents" or the like...

(I'm trying to piece the data together for myself, and I'm a bit suspicious that their graph stops in April. I'd also want to cross reference to number of vehicles on the road, because if there's been a ramp up since e.g. January, you would expect more incidents to happen).

I've just got to say I'm so impressed with both Waymo and Cruise. Last year they did 4.5 million miles put together in California with zero fatalities. This is an incredible achievement.
I will provide a dissenting voice. I was riding a Cruise with a friend the other night, leaving a bar - they only serve a nighttime crowd right now. suddenly a drunk hippie on Haight street started banging on the windows. Rather than get on our way and get us out of there like a Homan driver would, the cruise stopped, powered down - as in unlocked the doors, and called customer service. I shudder to think what would have happened if the hippie hadn't been drunk/inconsequentially high and was on PCP or something. This was a real thing that happened. I reported it to customer support. I can only imagine the conversations internally of this incident. (maybe I should get a job as an sre there.)
They include incidents where the car has caused traffic disruption. Recently, people have been throwing cones on top of the cars causing them to stop and impede traffic. This includes SF city officials, which is wild.

Look for #weekofcone on Twitter.

I feel like I'm being gaslit by the AV companies, especially Cruise. If these cars are so safe, why do I constantly seem them doing sketchy stuff on public roads? Why are there so many videos on Twitter and Reddit of pedestrians literally dodging out of the way to avoid being run over? It doesn't pass the smell test.
I am on my 3rd Tesla- been driving them since 2015- and frequently use FSD beta. I’ve watched autopilot develop into FSD.

The thing that baffles me about all current autonomous vehicle efforts is how resistant they are to having any sort of environmental feedback engineered specifically to help driverless cars.

For instance, in addition to flashing lights, what if fire trucks broadcast a “keep away” radio signal? Perhaps engineered specifically to be easy to determine distance and direction. Perhaps driverless car companies would have to pitch in to pay for the transponder refit for every fire truck. This is just an idea from a layperson and probably unworkable in this form but you should get the drift of my suggestion.

Also, reading this story, why didn’t the city place orange cones in addition to caution tape? I seriously doubt any driverless car company is training their vehicles on caution tape; as a driver for 40 years I have never encountered caution tape in a way to influence my driving. If the city is allowing driverless cars maybe they should have rudimentary training and standards for personnel to avoid unsafe situations.

Ditto for the firemen and fire hoses - ORANGE CONES are what tells me not to drive somewhere.

I worked as a software engineering tech lead on the "Ground Truth" team at GM Cruise for a short period in 2018.

During orientation, the founder (Kyle Vogt) told us all that the cars don't need to be better than a human driver, they only need to be "not worse" than a human driver. In fact, when questioned about it, he said that the sooner we could get robot cars on the streets of SF the sooner the software could improve.

The idea was that the more dangerous, "not worse than a human" era would be a necessary sacrifice of safety (compared to waiting forever until we had 100% perfect driving robot cars) -- so we could fast-forward to major improvements in the robot cars' capabilities.

That (IMO) cavalier attitude, a lack of rigor in the way software for the vehicles was being developed, and the fact that we had constant meetings about "diversity and inclusion" rather than on robustness, safety, ethics and quality pushed me to resign abruptly and go my own way.

Very disappointing experience - as I had hoped to see "The Future" before joining GM Cruise and even a fat paycheck and RSUs weren't enough to sooth my scruples.

Its okay if a few people die as long as Big Tech gets to gobble up more of our lives and make a few more billion in profits.
San Francisco won't be satisfied until they've completely driven out all the tech companies and their employees.
The incidents mentioned in the story would have been shared via first responder dispatchers. Seems like the way forward here would be for the city to make this data available to all driverless cars- maybe some sort of data registration with the city- and for the vendors to make it possible to drop these new hazards into the cars' maps promptly. I don't know how these systems work, or if they can work together, but I can write a computer program and there is clearly a looping input that is lacking in this system.
I really think incidents by autonomous cars are the least of SF's problems
My stance have shifted a bit lately, from "concerns/what ifs" to "let's see", I began to look at it through "the sooner the better" lenses.

Advancements like these are inevitable and have been on the radar for a long time. The more people familiar and informed about the tech, the better they are equipped in order to put the right regulations.

I'm definitely a skeptic when it comes to putting an AI in charge with no manual override of a 3 ton vehicle with humans on board, but having said that, this article didn't provide any meaningful information if these "incidents" per mile driven are higher compared with human drivers? They say they saw "huge increase in incidents". Perhaps there was a "huge" increase in miles driven by these cars?

It is possible likely hood of such incidents per car/mile increased due to a software update or tweaking of some parameters. I heard people often complain self driving cars drive too cautiously. Try to address this, it is pretty certain there will be more incidents. But is it what happened here? We have no idea.

In general I feel the quality of journalist output diminishes in time, or perhaps it was always horrible? How can one do research, spend time talking to people, write an article and don't provide basic numbers?

I’m not sure looking at the absolute number of incidents is very informative. How did the total number of miles-per-day from autonomous vehicles vary in the same timespan?
Hmm. I always thought these self-driving taxis were competing with regular taxis. But Cruise is saying "They also never drive distracted, drowsy or drunk," which makes me think they don't think they're competing with taxis, but with random drivers coming home from the bar at 2AM. I'm not sure I quite understand their argument here.
The easiest solution is to just pull Cruise's licenses because they've been and continue to be the worst offender -- the only reason they got significantly faster approval than Waymo to do things even more autonomously was because they're GM and leveraged their massive lobbying power to get away with breaking shit in production.
What's the problem that autonomous cars aim to solve? Almost everybody today is able to drive. Most people are safely navigating billions of kilometers everyday. People who are physically disabled or don't want to drive for whatever reason have the option to order an Uber. So what is it that will change if we attain a 100% human-like AI driver?
https://archive.li/jpVut
Driverless cars are probably safer in the US, because ppl here suck at driving :) In Germany you have to take driving lessons with professional instructors, including highway and night drives. The driving test in the US is ridiculously easy. You don't even need to be able to reverse-park.
I don't understand how these driverless cars don't have some sort of override for public safety officials to move them in an emergency. It seems like that would go a long way to reducing concern? Or is this just rabblerousing?
Now present the data in incidents per vehicle mile and compare it to human drivers please? Very difficult to take anything seriously that doesn't compare this technology to humans when it comes to safety. We're terrible drivers on the whole.
I miss Uber pool. It was $5 for a ride across town and it was a great way to meet people. Why do we need this self driving Taxi shit? Would much prefer my own car to have that functionality than to rent it out on a per-ride basis
Is this one of those "400% increase from 2 to 10" things?
At some point we will need separate emergency technologies to deal with automated vehicles as a fill in for exceptional cases. the situations described here are such cases
It's amazing after all the billions they've thrown into autonomous driving that it's still isn't a solved problem.

Must be an extremely hard problem.

It would be more useful to have some actual statistics comparing this to human drivers instead of scare pieces. Because accidents did happen before.