The self-driving future has arrived in San Francisco. And, increasingly and all too often, it looks like a confounded cop, road flare in hand, commanding a wayward autonomous vehicle as if it were a misbehaving, two-ton puppy.
“No!” shouts the cop, as captured in his body-worn camera footage. “You stay!”
The incident occurred on Feb. 9, during one of San Francisco’s more memorable recent emergencies: A dollar-store Walter White apparently lost control of his Sunset District garage dope factory, resulting in a lethal explosion and fire. The normally sedate neighborhood in the vicinity of 21st Avenue and Noriega Street was instantly transformed into both a disaster scene and a crime scene.
And, to make it a truly San Francisco scene, a driverless Waymo vehicle subsequently proceeded to meander into the middle of things, like an autonomous Mr. Magoo.
“It doesn’t know what to do!” shouts an officer caught in the background of the body-worn camera footage. “I’ll pop a flare!” responds the cop wearing the camera. “There’ll be hella smoke in the front.”
He then commands the vehicle to “stay,” and places a flare in front of it. But it does not obey, hella smoke and all (“ah, fuck!”).
“Got a bit of a pickle,” he radios to a dispatcher. “Got an autonomous vehicle, the Waymo, it’s inching slowly and closely to one of the main water lines that the SF Fire just charged. Can’t run it over. Is there a way you can contact a responder to come out and disable this vehicle? I don’t trust this AI.”
The cops eventually get the car shifted into “park” in the middle of the intersection. The officer, evidently a very positive person, makes the most of the situation. Taking a look at the autonomous Jaguar SUV that couldn’t recognize a fire hose and didn’t stop even after he ignited a road flare, he tells his cop buddies, “You know what? It worked out. Now, it’s like a barricade.”
And, to himself: “Self-driving. It’s where we’re at now. Technology.”
Asked for comment on their vehicles’ incursions into emergency scenes, Waymo provided this lengthy response.
Cruise wrote, “We are constantly improving our technology, and have a great relationship with the San Francisco Fire Department and other city stakeholders as we have deployed and expanded our service. We’re proud of our safety record and remain committed to doing everything possible to make roads safer.”
The “great relationship” part is probably true; fire officials who’ve worked with the autonomous vehicle companies had nice things to say about their representatives. But not about their vehicles.
“They’ve made every effort to work with us in public safety measures and be a good partner,” said one. “But they do not have a good product.”
Driverless Cruise and Waymo vehicles whirring through town have rapidly shifted from a novelty to a ubiquitous feature of San Francisco living. You’ve probably noticed this. Your emergency responders most certainly have, and their initial laughter at the cars’ antics has long since given way to irritation and trepidation.
Mission Local has obtained some 15 Fire Department incident reports documenting dangerous and/or nuisance situations in which Waymo or Cruise vehicles interfered with fire vehicles or emergency scenes. The vast majority of these reported incidents occurred in recent months, and a majority took place in April (driverless cars were only in December given the green light by the state to traverse San Francisco 24/7).
These incidents are either happening more regularly or being documented more regularly — or both. Within the marginalia of reports written last week, fire department officials complain that driverless car incursions are now a “daily occurrence.” This does not appear to be an overstatement: The notes on an April 26 report state, “This is an increasing problem. I believe there are many more incidents that are not being reported.” A subsequent note states “Number 3 today!”
“Before, you only used to have these things operating after 2 a.m. There’s not so much traffic; you go around them. It was funny then,” says a veteran firefighter. “Now they’re out in the daytime.”
This fireman’s observations jibed with observations from others Mission Local spoke to and with incidents documented in reports. Automated vehicles, he said, react strangely to flashing lights and sirens: “They just stop dead in the middle of the road. They freeze. I liken it to a deer in the headlights.”
And that gets to the heart of the problem right now. While computerized drivers are all but certainly more technically proficient than a human driver, the cars rolling around San Francisco currently can behave erratically, and in non-intuitive ways that throw emergency responders off.
“I shouldn’t have to look over my shoulder for a car trying to pierce a public safety scene that’s already been cordoned off,” says a longtime San Francisco firefighter. “This is happening more and more. Those cars can be relentless.”
That was clearly the case in late January on the 1300 block of Hayes Street. A firefighter reported that an “electric car with no driver” would not stop rolling into the fire scene and “was going to run over our hoses and possibly put our firefighters at risk.”
“I yelled at the car twice to stop, banging with my fist on the hood,” a firefighter wrote in the subsequent “Unusual Occurrence” report. “After warning car twice, I smashed the window and the vehicle stopped.”
Some of these incidents really do seem plucked from a Warner Bros. cartoon. Firefighters working in the wake of a March 21 windstorm report two Cruise vehicles rolling through warning tape and straight into the downed Muni wires the fire department was on-scene to deal with.
Then, like Wile. E. Coyote running into the man-sized sling-shot, the cars kept rolling until the tension of the wires entangled in their roof apparatus tightened to the point where they ceased driving.
“This incident raises many serious concerns about the safety of these Cruise driverless vehicles,” wrote the firefighter who filled out the report. “The need for these vehicles to recognize a road closed by caution tape, and caution sandwich boards is imperative. Secondly, the vehicle failed to recognize the large gauge Muni line hanging in its path. If this wire had still been ‘hot,’ this would have been much more hazardous. It is also of note that the vehicle did not recognize when it hit the heavy wire, or that it was being dragged on its roof top for half a block.”
redactedReports_roboCars_4 by Joe Eskenazi
To date, no firefighter has been run over, and no fire victim has suffered because emergency personnel have been unable to move their vehicles or access fire hydrants. But this, like the fact there was no electricity running through those downed Muni wires, appears to be a series of lucky breaks.
On the evening of April 26, a driverless Cruise vehicle stopped just behind a fire rig parked on Pine Street and would not move. Per the subsequent report, firefighters were forced to stop what they were doing and go about “poking and prodding the vehicle, and pounding on the windows until the driver’s window rolled down.”
A firefighter then spoke via a radio in the car to a Cruise employee. “The individual apologized for the ‘inconvenience,’ and said a team was working on moving the car,” reads the report. “Even after talking to him, the car did not immediately get moved from the scene. I informed him that this time it was an ‘inconvenience,’ but if someone needed to be rescued by the Truck it could have been a life and death situation.”
More than a year ago, Cruise worked with local fire and law-enforcement personnel to produce this 19-minute video instructing emergency responders on how to interact with its vehicles.
This video has all the flat line-readings of an early ’90s infomercial (RIP Don Lapre), but the high-quality camerawork and insane production budget of a 21st-century tech company. The video concludes with firefighters demonstrating how to saw through a disabled driverless vehicle without being electrocuted, and hacking a perfectly good vehicle into pieces.
Again and again, the video instructs emergency responders to call a phone number and defer to subject-matter specialists. The phone number is repeated with 1-877-KARS-FOR-KIDS regularity, and it’s hard not to sympathize with emergency responders now additionally saddled with making phone calls rather than, you know, responding to emergencies.
The increasing number of autonomous vehicles traversing San Francisco have generated recent local headlines for randomly stopping and triggering traffic jams; being unable to navigate the city’s ever-present fog and even slamming into an articulated Muni bus.
Every firefighter I spoke with wanted them off city streets to work out the kinks, posthaste. But that’s not the call of a local fire department — or, it turns out, any local official. Autonomous vehicles are regulated at a statewide level. And, in addition to tech wizards, all of these companies have government wranglers on the payroll.
That’s why their vehicles navigate through government regulation with so much more ease than they do the actual terrain of our city.
The next step, as it often is in San Francisco, appears to be legislative guerrilla warfare. A resolution approved by all 11 supervisors (but left unsigned by Mayor London Breed) concedes that local government, by and large, does not have permitting authority regarding autonomous vehicles. The resolution, however, states, that the city can indeed put a hand on the wheel (or a foot on the brake) via regulations on “potential operators using the public right-of-way, including but not limited to fleet charging, fleet deployment, curb management tools, tax incentives, fee waivers, and other approvals.”
So, that may be one version of our city’s future. Among many.
After handling the wayward Waymo at the exploding-house scene on Feb. 9, the officer blew off some steam with his colleagues. “The flare didn’t work!” he said with a laugh.
“I don’t know about you, but I wouldn’t get in front of a robot,” says a fellow cop. “It’s only a matter of time, dude. Have you guys seen ‘Terminator?’”
Uproarious laughter. “Skynet is awake!” More laughter.
And it was real funny. For now.
Map by Will Jarrett. Basemap from Mapbox.
Point #1: You all really have got to stop being so innumerate. Please look up the number of major injuries per thousand/hundred thousand/million miles driven by AVs and then compare that, changing denominator to miles driven by people.
Point #2: SFPD/SFFD needs better training to deal with AVs. It’s right there in the video. When you approach an AV, chock & block the wheels first. Those giant SFPD SUVs are perfectly capable of carrying those around. Then call the first responder hotline.
Point #3: This is just the same old tired litany of comments railing against ‘The Other.’ It’s the same car-addicted lizard brains who rail against bicyclists who run stop signs, take up too much space on the road, or get in their way when they approach an intersection (because they’re always driving well over the posted speed limit). It’s the same car-addicted lizard brains who barely even pay attention when yet another headline pops up about someone being killed either in or by a car. I guess the latter is understandable because it happens on average 10 times **a day** in the United States.
Will. Please sign up for a ride-along with the fire department. Preferably after midnight. You will not need raw data or “context.” You can see for yourself.
Tom, I’m replying to your comment because mine is not approved yet. But your comment is one of many that show y’all really need some data-interpretation lessons.
Re my comment below—whoopsie, now it is I who is innumerate! I meant to say, 100 TIMES A DAY. One. Hundred. People. A. Day. Killed by or in a motor vehicle crash. In the United States. Compare that with pretty much any country in Europe—one of the few places in the world that are embracing the future of EVs and AVs but also trying to get cars out of their cities altogether. We need more automation, not less. Speed cameras. More red-light cameras. Speed governors on vehicles. Better traffic signals. All of you are squawking like Chicken Little when you can’t even see the real monster.
It’s a crime to willfully refuse or fail to comply with a lawful order or direction given by a peace officer under the California vehicle code. The corporations who could benefit from risking public safety, should also be fined for when their vehicles fail to follow the same laws that humans have to follow.
Article uses the type of anecdotal / no context data references that Mission Local has been critical of in San Francisco crime debate
All the data and all the reports are embedded. They’re not “anecdotes.”
Raw data counts equals “no context”, IMHO, as no there is no attempt at an appropriate baseline for comparison. Similarly, an article about X murders year to date in San Francisco [with embeds to police reports, a map of all the murders, and anecdotes from police at murder scene investigations] with no mention of the murder rate in SF compared to other cities would be missing crucial context as well.
The baseline of incidents with driverless cars can be safely assumed to be zero prior to the recent invention and deployment of driverless cars. Hope this helps.
How many cops does it take to protect a fire hose from a Waymo? I’ll stick to human drivers thank you. Wonder how they handle speed bumps and humps? Do they slow down or is it just a bumpy ride?
Maybe there is a good answer to the following situation, but in all the articles I’ve read about driverless vehicles, I’ve never seen it addressed.
Suppose you are in a driverless vehicle taking you from Point A to Point B, and something unexpected comes up. What do you do? For example, years ago I was taking a taxi to the airport. About four blocks from home I realized I had forgotten something important, something I had to have with me. I told the driver to turn around and go back, I had to get something I forgot. No problem. Went back home, I ran inside, got what I needed, and still got to the airport in time. What would happen in that situation in a Waymo or a Cruise? So far at least, I think I’d rather stay with my fellow humans.
That is some top notch local coverage and delightful writing.
So the tech execs who are afraid to walk a mile in San Francisco consider our much-maligned little seaside town of lunatics, criminals, addicts, corrupt officials and homeless hippies to be “real life”. Take that L.A.!! But one question: Is anyone driving these companies?
You’ll never look at autonomous vehicles the same way after reading the novel Attack Surface by Corey Doctorow, which with eerie synchronicity I just finished. In the just-a-few-years- from-now Bay Area, the Oakland cops have become increasingly militarized, and weaponize autonomous cars to kill protesters. Utterly chilling — and far from impossible.
I just don’t understand what is the problem these cars are solving?
The self-driving part of the equation kind of makes sense to me: sometimes computers can do things better than people, and maybe driving is one of them. I think it’s possible (although still not proven) that self-driving cars will reduce accidents and increase pedestrian/cyclist safety. That would be amazing.
But the autonomous part makes no sense to me. Why do we need to have cars without people in them? Are we experiencing a people shortage such that our true potential is not being reached because there aren’t enough people to drive all the cars we want to have out there?
I’d be much more interested and open to self-driving cars if this tech was truly about increasing safety and eliminating pedestrian and cyclist harm caused by cars. Get it working well and then make it mandatory in all cars. Program them to follow all speed limits, red lights, yellow lights, stop signs, yield signs, crosswalks and bike lanes (and to not do donuts ever!!), etc. But still require people to be in the cars. The robot would do its job improving the safety of our streets, while the “driver” would be there to deal with edge cases and emergencies that the robot can’t handle.
It costs money to pay people to drive things (and other people) around, so I think one of the ultimate goals is financial efficiency by taking human labor out of the equation. This would ultimately include delivery drivers and even long haul truckers. I don’t think this is in the greater interest of society.
“It costs money to pay people to drive things (and other people) around, so I think one of the ultimate goals is financial efficiency by taking human labor out of the equation. This would ultimately include delivery drivers and even long haul truckers.”
Sure, sure, all this improved “efficiency,” but then when the lack of jobs turns into more “lunatics, criminals, addicts, homeless” invading your neighborhoods, it becomes apparent all this “efficiency” is more of a theoretical solution looking for real-world problems, exactly like those inane cars driving around with nobody in them. In other words that’s not a solution, it’s some tech CEO trying to make his second billion.
“I just don’t understand what is the problem these cars are solving?”
The immediate problem they are solving is liberating scores of billions of dollars of cheap loans, early-round
speculationinvestment, and security/military state black budgets.
There are two main long-term “problems”:
1. Putting long-haul truck drivers out of work, because the peasants must be crushed;
2. Shoring up dwindling military recruitment with autonomous tanks and other military vehicles. Autonomous vehicles eliminates the time and expense of training human capital on complex equipment, bodies that will likely be on the job for only five or six years, assuming they aren’t blown to bits in battle.
AVs aren’t viable as a retail product, as they will never be affordable off-the-lot for anyone who can’t already afford a chauffeur. They will be also too expensive to buy/maintain/service/upgrade/replace for the livery arbitrage pirates uber and lyft to be feasible.
I wish people would stop using the term “AI” to describe these systems. There is NOTHING intelligent about them. They are just statistical models.
If you or I ignored the instructions of a police officer and continued to drive towards them in a car, the best case scenario is we’d end up in jail. But it’s also likely we’d end up dead. What’s the equivalent for an AI car corporation?
Given the nature of humans, predictably unpredictable, irrational, nuanced, and less than obedient, distracted, reactionary, and of variable speeds, the versatility required for these AI systems to operate safely in such an environment with a level of accuracy that supersedes our concerns is daunting. I’m grateful to know one such tech operator works for a company not mentioned, because the level of service they are trying to achieve in dense, diverse, and erratic sample population requires constant review and intense learning curves in order to deliver the safest fleet out there. I wish most drivers could be as focused, capable of looking out for the good of society, and efficient, rendering such driverless vehicles unnecessary, but the biggest argument for them is our own lack of care and consistency in operating personal machines.
Without self-regulation, as in other hotly contested targeted topics, more external regulations pile on top. At least the automated drivers won’t shoot at you. Yet. Hope they add to their learning curve regularly, as humans tend to forget everything they needed to know to pass the test, once they are licensed. Be good humans, look out for each other, and realize these machines are not going away. The City still operates cable cars and other relics of mass transportation. There’s so little parking left anywhere, and many who will not be able to drive themselves anywhere stand to benefit from a well-maintained and regulated autonomous vehicle operating system. Hope they work out the kinks.
“More external regulations pile on top”? So you think these monumentally-wealthy companies should continue to use the city as a giant lab for testing their dangerous robots, to the detriment of its citizens, all while being far less-liable for infractions and poor performance than a human driver is? A human driver that drives like this over and over would get their license pulled and possibly thrown in jail, they don’t get to just say “sorry for the inconvenience – It’s a learning experience!” and go about the same behavior.
I bet you’d see better behavior emphasized in these cars right quick if the company executives were held liable like human drivers for the dangerous infractions their machines rack up.
A Cruise robot car had traffic completely blocked on Kezar near MLK this morning. Looked like it freaked out because of construction going on in the middle of the street which blocked one of the lanes.
I wave at the heartless vehicle like it’s my best friend. I hope some engineers understand that’s my safe way to give it the finger. Stay sane.
My Roomba robot vacuum cleaner came with devices that create an invisible signal beam that blocks the robot from crossing the line. You’d think these driverless car operations would equip first responders with similar devices to create a kind of electronic yellow crime scene tape that would block robot vehicles from entering. But hey, I’m no genius.
Good suggestion. If flares won’t do, yelling is futile, digital age requires digital solutions.
Really stupid idea. Never would any driverless car operation cover such expenses, this is for taxpayers to cover. /s
Seems like a geo-fence around emergency responders would be a good idea. The fire and police have an app or a pre programmed key on their trucks, ambulances, or cruisers and the hit it when they roll up on an incident that needs to be free of robot cars. It is imperative to keep in mind that these vehicles hold a lot of promise when it comes to driving down the fatality rate. If we currently have 25000 fatalities and these robots can get us to 10000 fatalities then we would be remiss not to employ it. We can’t let the perfect be the enemy of improvement.
The technology is not yet at a point that it is safe and reasonable to have a car operating with no one in it. It may or may not get to that point, but it’s absurd that it is allowed now.
That is exactly what I think.
Autonomous driving is not your conventional engineering problem, where (you hope) only a relatively small number of edge cases need to be covered. BART was conceived as and would have been capable of a fully automated operation, because BART runs in a controlled environment. Autonomous driving OTOH is all edge case, impossible to cover with an engineering approach. So the idea is for AI to fill in the blanks. Where the wheels come off though is with a technology that is limited to generalization (ANNs). There is no capability for abstract conceptualization, for learning, for inference, regardless of the language often used by practitioners. And that’s needed to pick up on abstract barriers like traffic cones lined up further than a car’s length, or caution tape run across the street. It is needed to navigate a dump truck situation as described in the comments. And on and on it goes, a completely open ended problem, which stands no chance of solving any time soon, because an actually intelligent AI is, for decades now, just decades away. And yes, this has played out in academia since the 50s, devoid of real world implications as we see now with Waymo and Cruise (and chatbots).
I’d venture Cruise and Waymo just have been lucky thus far nobody was reported injured as a result of their operations. Only a matter of time though, unfortunately.
I am a 5th generation San Franciscan, and in my 51 (oy!!) years living in this great city I have experienced many great and tragic things. However it has only been in the last few (maybe 10) years in which I have been forced to wrestle with, on a human level, the stark and obvious insignificance we have as individuals (with our supposed rights to be free from tyranny) when compared to the interests of the disproportionately wealthy, and therefore powerful, corporate entities that represent the, very much for-profit, AI industry. There can be no doubt that none of us have been asked if we are willing to offer ourselves up as living, breathing and feeling moving targets upon which these venture class, already wealthy, business interests can hone their products in a “real life” environment. Perhaps it is important to not forget that many of these technologies are of great interest to the military industrial complex and are ,no doubt,funded, in part, by them as well. My fellow San Franciscans, I ask you: Were you consulted? Did you give your permission to allow our home; where our children play; where our loved ones walk the streets and our elderly folks take their well deservedly sweet-time crossing the intersections, to be the testing ground for these fast moving, heavy metal objects?
Unless I just missed it, the answer is an echoing no.
I have heard all of the arguments regarding how this technology may have a helpful impact in our future, and that is all very well. I have also spoken to, and with, a number of people who work within this industry,and whilst many of them have the zeal of the truly converted, to a one, they, in my opinion, far too easily wave away the concerns of the individual and those of us who have no real voice because we are not Uber wealthy, in favor of their chosen future. This very much reminds me of the young person who, despite the opinions of those with more experience, refuses to listen because it would mean they have to slow down and, God forbid, admit they may have rushed forward on a faulty premise and have honestly made a mistake. Now imagine how much harder that would be if your efforts and intelligence were being rewarded by both grossly large amounts of money and compliments to your youngish ego. How could this be wrong? Look!! I was able to buy a multi-million dollar home for my family in one of the most expensive cities in the world!! I must be doing something right.
But. But, given the both foreseen and experienced,and glaringly unforeseen, tyranny of the reality and the possible outcomes of the near future tyranny this promises to unleash on humanity, I, in my humble opinion, would say that the time is past to protect the fragile egos of the recently ascended mighty and wealthy in favor of us, The Everyone Else. Who’s interests, both economically and physically, have been waved away in their consultations and actions up till now. Tyranny comes in many forms, even cloaked as success, or as progress, and if we do not stand up and voice our opposition to it, then, my friends, we welcome it, and shove the door wide open to it in the future because we allowed it in the first instance. Sorry, but here we are.
Saw one come to a complete stop on 2nd Street after a recent nighttime Giants game. After reading this article I’m guessing it was confused by the flashing lights on nearby police motorcycles that were blocking some side streets? Other cars, taxis, etc were clearly frustrated being stuck behind it and the pedestrian crowd mostly laughed & took photos & videos.
Report with video/photos all naughty driverless cars to AVFeedback@dmv.ca.gov
Great research work here Joe – I knew a photo of the Cruise with a broken window existed somewhere, and you managed to find it. Such a great representation of what’s wrong with AVs today.
It’s not just emergency vehicles. I watched a Waymo stare down a giant garbage truck on a narrow street. The Waymo was in the center of the road, and there wasn’t room for both to move. A real car would have pulled into a parking space to let the garbage truck by; that option was there. But the Waymo refused to move, so eventually the garbage truck had to back up and go down another street. I suppose that garbage was not collected that day.
Artificial intelligence and the “self-driving” cars it controls don’t and can’t actually think; what they do is use probability to try to match a novel situation with identified patterns in its data set. AI would be more accurately termed “GI”: genuine stupidity.
What can go wrogn, or even wrong, will.
I wish Mission Local would stop using the terms “guerrilla” and “guerrilla warfare.” It’s sloppy shorthand which adds a ton of bias and connotation. From Brittanica: “Over the centuries the practitioners of guerrilla warfare have been called rebels, irregulars, insurgents, partisans, and mercenaries. Frustrated military commanders have consistently damned them as barbarians, savages, terrorists, brigands, outlaws, and bandits.”
The average citizen doesn’t want to get involved in anything to do with “guerrilla warfare.” Using this term only serves to further scare the public away from productive civil engagement which is necessary for a functional democracy.
Or we could just consider actual guerrilla warfare, and recognize that the tech overlords of SF are already fighting asymmetrically as the equivalent of an unaccountable, entrenched caste with military (government) support. Last time I checked guerrillas are generally viewed more-positively than oppressive regimes by most people, unless you are a big fan of oppressive oligarchies.
I agree with you 100%.
I’ve found a couple of them paralyzed/confused by the construction on 18th street these last 4 weeks