Self-Driving Uber Car Kills Arizona Woman

Your life means nothing if it conflicts with maximal shareholder value.

A woman in Tempe, Ariz., has died after being hit by a self-driving car operated by Uber, in what appears to be the first known death of a pedestrian struck by an autonomous vehicle on a public road.

The Uber vehicle was in autonomous mode with a human safety driver at the wheel when it struck the woman, who was crossing the street outside of a crosswalk, the Tempe police said in a statement. The episode happened on Sunday around 10 p.m. The woman was not publicly identified.

This notion that having a "safety driver" in the passenger seat will allow a distracted human to take over at the last minute is completely insane. You think driving-while-texting is dangerous? This is so much worse. When people aren't engaged in the task of driving, their minds wander. They cannot re-engage fast enough. This is obvious on its face, we don't need studies to prove it. Oh, but we have them anyway.

Uber said it had suspended testing of its self-driving cars in Tempe, Pittsburgh, San Francisco and Toronto.

"Our hearts go out to the victim's family. We are fully cooperating with local authorities in their investigation of this incident," an Uber spokeswoman, Sarah Abboud, said in a statement.

"Thoughts and prayers, but not one dime."

Autonomous cars are expected to ultimately be safer than human drivers, because they don't get distracted and always observe traffic laws.

Nice, weasely use of the passive voice there, New York Times. Expected by whom? Certainly not by anyone with any expertise in computer science. Or AI. Or anyone who has ever used a computer. Or a cell phone. Or a printer. Or driven a car.

However, researchers working on the technology have struggled with how to teach the autonomous systems to adjust for unpredictable human driving or behavior.

"No 9000 computer has ever made a mistake or distorted information. We are all, by any practical definition of the words, foolproof and incapable of error."

"It can only be attributable to human error."

Note that the article's headline referred to the woman killed by the robot as a "pedestrian" instead of a person. "Pedestrian" is a propaganda term invented by the auto industry to re-frame the debate: to get you to preemptively agree that roads, and by extension cities, are for cars, and any non-car-based use is "other", is some kind of special-case interloper. See The Invention of Jaywalking.

Semantics aside, I have one question that I think is pretty important here, and that is, who is getting charged with vehicular homicide? Even if they are ultimately ruled to be not at fault, what name goes on the court docket? Is it:

  • The Uber employee "non-employee independent contractor" in the passenger seat?
  • Their shift lead?
  • Travis Kalanick?
  • The author(s) of the (proprietary, un-auditable) software?
  • The "corporate person" known as Uber?

Previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously, previously.

Tags: , , , , , , , , ,

43 Responses:

  1. Thomas Lord says:

    Oh, god. I'm four years too late in reading one of the "previouslies" but, man, that parking-sandwich analogy is trash.

      • Thomas Lord says:

        Just to be clear, I think this kind of life is preferable to dying at the hands of a car in a society that is always rushing everywhere:

        but the sandwich analogy for on-site parking requirements is very wrong economically and also overlooks the (both realized and potential) non-parking social uses of on-site parking spaces (which, after all, are just space, that can be used, in various ways).

        • jack lecou says:

          What's 'very wrong' about it? It seems to get the gist of the problem across well enough to me. If anything, it's soft-pedaling it a little: to be complete, you'd have to mention the follow-on distortions beyond just the subsidy, like how a plethora of free/cheap sandwiches disrupts the supply of other foods, etc.

          And, I'm sorry, but "non-parking social uses"?

          I'll grant you that disused parking lots might occasionally play host to something like a farmer's market, street hockey game, barbecue stand or carnival tent, but nevertheless I strongly suspect that about 99% of the vacant parking spots out there are simply vacant 99% of the time. The emphasis is very much on unrealized potential there, such as it is.

          Parking spaces are, after all, not just space. They are space which has been laid out and structured in a specific way for a specific use - graded for runoff, paved with a hard, heat absorbing surface, delineated and interrupted by bollards and curbs, lamp posts, access ramps, etc. All of that makes for a very poor -- even somewhat hostile and forbidding -- substitute for the kind of public space you'd design for any given (or general) "social use" by actual human beings. It's no surprise they aren't much used for that.

          And that's ignoring that the highest value alternative use -- i.e., the opportunity cost of maintaining that parking space -- isn't going to be an equivalent amount of more appropriately open space, but other actual structure for specific purposes. Housing, shops and restaurants, etc. Without all the [often empty] parking, those things would be denser and closer together, and/or less expensive to build.

          • Thomas Lord says:

            The sandwich analogy describes an ongoing, real operating expense ($10 per day per resident). The operating expenses of residential parking are approximately $0. The $30/month gross income the landlord gets for some of the spaces (in the example) is almost entirely net income.

            The total net income of real property and the general rate of interest together determine the price of the real property. As a rule of thumb, divide the net income by .05 or .06 to get a sense of the property value.

            That property value, based on the net income, is accounted partly as the improvements and partly as land. This division is not arbitrary. The current costs of building equivalent improvements is externally imposed. That amount, subtracted from the total property value, is the land value.

            Thus, the incidence of a required construction cost -- parking, hooking up the sewer system, making an adequate foundation, etc. -- is on the land owner. Those $3 sandwiches aren't what makes the rent too damn high.

            You asked what I meant by "non-parking social uses":

            Real world examples near where I live: parking used as play areas; covered parking used as a shared cache of bulk-purchased commodities; more than a few wood-shops.

            If you find yourself in a situation with lots of protected, un-used, spaces consider their utility for things like a storage area, pool table, speakeasy, lounge, commerce, etc. This is where I would rather see regulatory reform go -- not towards building lots of parking-free housing, but making it easier for tenants to control those spaces and use them in non-traditional ways.

            I also hold the unpopular opinion, by the way, that while there needs to be a massive, massive reduction in the amount of driving that goes on, it is a big win for public safety / resilience to have lots and lots of working cars distributed throughout urban environments. This is another reason to preserve some level of parking minimum requirements.

            • jack lecou says:

              The operating expenses of residential parking are approximately $0. The $30/month gross income the landlord gets for some of the spaces (in the example) is almost entirely net income.

              If you like. And if the sandwich-making landlord bought a 15 year supply of bread and cured ham upfront, I guess you could argue their 'ongoing real operating costs' were 'approximately $0' too -- but it wouldn't really change the argument at all. (FWIW, the fixed construction costs for parking are somewhere in the neighborhood of $24,000 to $34,000 per space, depending on surface lot/below ground, varying by region etc. That's a lot of ham.)

              All else equal, the money a developer doesn't have to pay to build out unused parking is going to translate directly into either being able to build an equivalent number of additional housing/commercial units on the site, or lower construction costs, either of which should translate to downward pressure on overall housing prices.

              And of course, the economic problem here is not really about the landlord's revenue or operating costs at all -- which is what you seem to be focusing on. It's about the distortions of the residents' behavior due to the cross-subsidy. For example, suppose some people are sandwich lovers and others are pizza lovers. Without the subsidy, both foods are about the same price, so the sandwich people get sandwiches and the pizza people get pizza. Neither food is given preferential treatment.

              Under the subsidy, however, things are radically different. The sandwich lovers are probably eating MORE sandwiches than they otherwise would, because the marginal costs are hidden (thus they're spending more on sandwiches than they actually want to). A bunch of pizza lovers are grimacing and buying sandwiches instead, because they can't justify paying more than twice as much for pizza (and they're getting charged for part of the sandwich either way), and the remaining pizza eaters are getting double charged, because they're effectively paying for a sandwich they don't eat but still shelling out for a slice.

              And the supply chain is probably getting f'ed up too -- pizza parlors going out of business while lunch meat suppliers are booming. It'd just be a mess all around, which as analogy for parking subsidies, is not bad at all.

              ...a storage area, pool table, speakeasy, lounge, commerce, etc.

              Yep. More (underground) utility spaces, retail spaces, recreation spaces etc. are exactly the sort of things that might be viable additions to construction or renovation projects if parking requirements weren't a thing.

              Which is what it looks like you're talking about. After all, you can't (legally) convert a parking garage into a speakeasy or a workshop if those spaces are serving to meet someone's minimum parking requirement. If you've leveled the pavement, put up walls and lights and air conditioning in order to turn it into a pool hall, it's not a parking lot anymore.

              Now, if what you're saying is that we should work on regulations or other incentives to encourage building, say, multi-use workshop spaces into residential buildings, I'm with you. But minimum parking requirements are just about the exact opposite of that...

              • Thomas Lord says:

                if the sandwich-making landlord bought a 15 year supply of bread and cured ham upfront, I guess you could argue their 'ongoing real operating costs' were 'approximately $0' too

                See, no, you couldn't. Not if you know what you're talking about. Cluestick: the ingredients are consumed in the making of the sandwich.

                FWIW, the fixed construction costs for parking are somewhere in the neighborhood of $24,000 to $34,000 per space, depending on surface lot/below ground, varying by region etc. That's a lot of ham.

                Which is why parking minimums lower land prices.

                Listen, you seem earnestly interested in reciting the entire book of YIMBY nonsense poetry. There's probably an abandoned tunnel somewhere that's a better place to do that. Sorry I tried to help you understand something. My bad.

                • jack lecou says:

                  See, no, you couldn't. Not if you know what you're talking about. Cluestick: the ingredients are consumed in the making of the sandwich.

                  cluepachyderm: that's not important to the analogy. Not any more than whether the bread is whole wheat or not. The precise nature of the inputs or the the fact that average (construction*) costs drop over the long run is incidental to the distortions created by the subsidy. The fact that it's land and not some more temporary flow that's being misallocated is important inasmuch as land uses and distortions of built infrastructure are much more difficult to correct, but the basic economics and common sense points the analogy is trying to get across are the same.

                  * Construction cost isn't the the only cost here anyway: the opportunity cost is what we're really concerned with. That score continues to run up for as long as the parking exists, and since parking is just about the lowest value use of urban land there is -- sometimes 20 or 30 times less than other uses -- that's also a rather considerable quantity of ham.

                  Which is why parking minimums lower land prices.

                  It seems like you've been gesturing at some other kind of argument based on real estate appraisal or something*, but if you actually think it's important, you'll have to refresh me on what it is exactly, and why it's relevant.

                  Because I'm very skeptical that it is. The blunt fact is that parking minimums create additional costs and reduce housing supply. We could also mandate that 250' tall nude statues of Steve Jobs be constructed every 20 yards. Regardless of their specific effect on land prices or returns to development investment, it hardly seems like good public policy. And yet the usefulness of the statues and their economic effect would be about the same as empty parking spots: simultaneously increasing the cost of construction and reducing the total supply of housing in a given region. (Actually, the statues are probably better. They reduce density, but at least they don't induce any externalities from transportation mode share distortions.)

                  * It looks like maybe something about capitalization rate -- i.e., that:

                  land price + construction cost = income / cap rate

                  ...and therefore, if construction costs increase, land prices must decrease correspondingly. The problem I see there is 1) what does it have to do with the price of tea in china? -- economically it doesn't really have much to say on the question of whether the subsidy affects density or is fair or good policy. And 2) that it's not at all clear that the right hand side -- income, in particular -- would necessarily be fixed in the presence of changes in regulations - a binding MPR regulation is a tax on construction, but it's also a tax on ongoing operations, inasmuch as those extra parking spots were, by definition, not profit maximizing.

                  • Thomas Lord says:

                    Jack, I come here for the pasties and the music industry insight and the nightclub point of view and the general weirdness and sometimes I try to add to the geeky fun with some Real Life Knowledge like here but if you want to take my introduction to real estate economics course to work through your issues it meets in Berkeley at ad hoc intervals in disreputable locations and the tuition is payable in Bloody Marys and whiskey drinks.

                  • jack lecou says:

                    I think I'm covered on the economics front, thanks. But if I'm ever in the neighborhood and spot one of those converted parking ramp speakeasies, I'll be sure to stop by.

  2. MattyJ says:

    "... unpredictable human driving and behavior" is pretty much all human driving and behavior. I ride a motorcycle often, and yes, I sometimes lane split, and this shit creeps me out. Robot cars should be required to have a big orange cone on the roof so people can identify them from 100 yards away. Can't wait until these robot flying taxis start falling out of the sky.

    PS That robot smashing the egg is my all-time favorite gif, hands down.

  3. Doctor Memory says:

    "Semantics aside, I have one question that I think is pretty important here, and that is, who is getting charged with vehicular homicide?"

    E. Nobody. Because nobody ever gets charged with vehicular homicide. In most major cities there is literally nothing -- up to and including killing a three year old while she was crossing in the crosswalk, with the signal, holding her grandmother's hand that will get you so much as a permanently suspended license nevermind a criminal charge. It is 100% legal to kill anyone you want with your car. Knock yourself out. Seriously.

    (Our friendly host, I imagine, is completely aware of this. But the Allison Liao case deserves to be better known because in retrospect it was vehicular homicide's Sandy Hook moment: the worst possible thing happened, the victim was a child, the method and means were 100% within the state's ability to fix, and literally everyone in any position to do anything about it from the officer on the scene on up through the Governor decided that the proper response was absolutely nothing. Six months later the driver's tickets were voided by the traffic court.)

  4. DWidel says:

    The stories all say Ped, but the pics show a mangled bike. Not sure there's a lot of difference between a pedestrian outside a crosswalk and a biker from the point of view of a robo car. Either way, not sure how they expect this to work with the technology we have.

    I'm also quite quite sure they will be safer than human drivers, I just don't think it's going to be in my lifetime, or anytime soon.

    • Yeah, it's still unclear exactly what the victim was doing, and where. I saw some claims that she was walking her bike, possibly in the bike lane there. Guess we'll find out once Uber releases the dash-cam video. But wait, it's Uber, the company whose entire premise is criminal, so don't be surprised if the video turns out to be lost.

  5. Yuma Tripp says:

    Killed by a robot car. It begins...

  6. Tim says:

    When they said "move fast and break things", I naively failed to realise that could apply to people, too.

    Oblig. xkcd.

  7. Disrupt crosswalks

  8. Tempe police say she was walking her bike in the street. I didn't see them mention the bike lane - there is one at that location.

    They also said the vehicle was going approximately 40MPH. Speed limit there is 35MPH. What was that about robocars will always obey the law?

  9. Sam says:


  10. ennui says:

    Tempe police appear to already be extending cop car privileges to Uber, saying the victim walked out in front of the Volvo and was probably homeless anyway...

    the hacknews thread is truly something. consensus seems to be that you have a moral obligation to kill people with your engineering projects, as long as the kill rate is less than some predetermined "normal" amount. of course the only way to determine the kill rate of your project is to test it in public. by the way, why exactly are they testing these cars in public?

    it gets back to the whether anyone actually believes pattern recognition on images (in whatever spectrums) via neural nets is the right technology for self-driving cars or whether the whole thing is driven by a) Google's endless need to convince the markets that all of the terabytes of data they are hoarding are actually just about to be incredibly valuable once the insert tech starts working and b) keeping Uber's grand pump and dump scheme going for another couple of rounds.

  11. Aracan says:

    Does the "driver" actually sit on the passenger side? What is the rationale behind not having him sit at the wheel?

    • jwz says:

      I assume the meat-drone actually sits in the so-called "driver" seat, but by my reckoning, someone who is not driving is a passenger.

      You know that dude's on his phone, too.

      • margaret says:

        You know that dude's not going to be one of Uber's allies when it's time for someone to take the fall.

        The driver, Rafael Vasquez, 44, served time in prison for armed robbery and other charges in the early 2000s, according to Arizona prison and Maricopa County Superior Court records. Uber declined to comment on Vasquez’s criminal record.

        Bloomberg's articles on this have been much less ¯\_(ツ)_/¯ than some of the other outlets but I'll be damned if I can figure out how to link into their news-feed.

      • Now that the dashcam footage is out, two things:

        1. It's a woman, so I expect the pro-robot folks to start arguing that if it was a dude they would have avoided this, and

        2. Whether she was the nominal driver or not, she was not paying any attention to the road. In the in-car video, you can see her startle immediately before the video ends (just before the collision, if it runs in parallel to the external video), far too late to do anything even if she had acted perfectly from that point on. Prior to that, she's looking around outside, and looks like she's daydreaming.

        Which is not to say that this collision is her fault. It's not, I don't think. The aesthetics and affordances of this particular technology specifically encourage drivers to trust the machine and let go of the responsibility. As you say, she's clearly a passenger, even if she's seated in the driver's seat. Even professionally trained pilots have trouble reacting appropriately when their automation cuts out abruptly; untrained lay people are almost guaranteed to be worse than that. In short, the position that the driver is responsible for taking over in an emergency is a design choice that contributed directly to this death: that woman was in an impossible position in part because of choices made by the designers of the system.

  12. ennui says:

    the solution to the liability problem is, obviously, to set up a market for sacrificial "drivers," who would sit in the drivers seat and assume liability for any accidents during operation. oh wait, uber and lyft are already doing that. but there's no reason why my insurance-disruption idea couldn't work for other industries.

  13. Not Frank says:

    Naturally, I'm betting all the other folks working on self-driving cars are glad it was Uber who killed the first pedestrian, or, per some of the reports, bicyclist.

  14. o.o says:

    I saw this yesterday and started to get wound up about it, but then I realized this is only news because it involved a self-driving car; any "regular" pedestrian death would only make local news, if that. What is the aggregate deaths per miles driven for autonomous cars vs. human drivers? Who gets charged is an interesting question, though.

    • margaret says:

      pedestrians are always poor people. killing poor people is never worthy of attention or action.

      carry on. nothing to see here.

    • Tolomea says:

      Yeah, that's roughly why I disagree with JWZ appearing to hate on the whole concept of autonomous cars. As an engineer with a lot of tech negativity and a fondness for saying "my industry has a lot to answer for" I still thing autonomous cars are a winner. It's important not to confuse perfect driving with better than average human driving. That said liability is very important it's a mechanism to push improvement, if there's no consequence for running over a pedestrian then there is also no motive to fix whatever happened.

        • ennui says:

          how many vehicular homicides get a prejudicial public comment from the police chief?

          but again, these public tests are primarily a PR and stock promotion effort. state-of-the-art for these cars is somewhere around level 3, arguably 4, but putting them on the roads conveys the impression (contra official announcements) that they are around level 5, just need to work out some kinks and are about to go into production. they still need a major advance in technology to get anywhere near production, which makes driving them around in semi-uncontrolled environments ie. public roads, not really engineering best practice: are they really getting testable, repeatable data from their trips?

          but because it's impressive to see any kind of automated driving, putting the cars on the road generates tremendous "buzz" for the investors involved. Google and Uber are willing to kill people if they ultimately promote the idea that their investors are going to end up controlling this tech, regardless of whether it's anywhere near working or actually working at all. the perception of a new field of tech that is tied into Alphabet stock or whatever is more important than actually having tech near production and certainly more important than an accidental death or two...

          moral panic over who is liable just distracts from the fact that there is little real engineering justification for having pedestrians anywhere near these cars right now.

        • jwz says:

          The fact that you are still able to mentally separate "this is a neat technical hack" from "hmm, oh yeah, corporate liability, that's kind of an issue, huh" is literally everything that is wrong with the tech industry.

          The way you are thinking? Or failing to be thinking? This is how we get Cambridge Analytica.

          But I'm sure "someone else" will work out that handwavey other half of the problem. Sure.

    • Honestly, this is an excellent opportunity to restart the discussion on how we treat vehicular deaths generally, and I'm sad that we're likely to waste it on "well, robot cars kill fewer people, so that's good enough." The way we design roads and cars is lethal, and putting a robot behind the wheel doesn't really change that.

  15. k3ninho says:

    cstross, of course offers a view from the 'dim shell-company of shell-companies registered in an off-shore pirate hell-hole jurisdiction where you could never serve a case' future (not the grim meat-hook, but let's call it a cousin): test-case at


  16. ennui says:

    and they release a low-res video, it's uber footage but the cops releasing it:

    a) can you imagine? the police chief seems to have misrepresented what clearly is a woman in the middle of crossing a well-lit road with a bicycle as "she jumped out of nowhere"

    b) the backup driver is... looking at his/her phone.

    c) this is pretty clear level 3 driving. ask yourself: why is this tech project actually on a public road?

  17. mr_mercer says:

    If I buy myself a dazzle suit and wear dazzle facepaint can I be reasonably certain of rolling off a hood and collecting $10mm in Tempe next month?

  18. Uber’s Self-Driving Cars Were Struggling Before Arizona Crash

    Average miles before human intervention required:
    Google/Waymo: 5600
    Cruise: 1200
    Uber: 13

  19. Nia Psaka says:

    Wow! I, not being particularly familiar with this blog, expected the typical techie line that yes, statistically, the machine will kill fewer humans than the human. The internet is full of that sort of talk.

    Thank you for the points in this post.

  • Previously