By on March 20, 2018

Uber Volvo Autonomous

Details are trickling in about the fatal incident in Tempe, Arizona, where an autonomous Uber collided with a pedestrian earlier this week. While a true assessment of the situation is ongoing, the city’s police department seems ready to absolve the company of any wrongdoing.

“The driver said it was like a flash, the person walked out in front of them,” explained Tempe police chief Sylvia Moir. “His first alert to the collision was the sound of the collision.”

This claim leaves us with more questions than answers. Research suggests autonomous driving aids lull people into complacency, dulling the senses and slowing reaction times. But most self-driving hardware, including Uber’s, uses lidar that can functionally see in pitch black conditions. Even if the driver could not see the woman crossing the street (there were streetlights), the vehicle should have picked her out clear as day.

The location of the accident further complicates things. Using information provided by local authorities and images from the scene, we found the the area where the accident took place. The location is a point in the road (northbound Mill Ave.) where a designated bike lane opens up to meet with traffic attempting to make a right turn. Following the incident, the self-driving SUV appears to have stopped in that turn lane with damage visible on the right side of its hood.

However, early reports (written prior to examining the car’s video data) claimed 49-year-old Elaine Herzberg was pushing a bicycle down the center median and crossed into the street when she was struck. There is a walking path there, which includes two pedestrian exits into to the street with a confusing “do not cross” sign. But how the damage came to be on the right side of the vehicle if Herzberg entered from the western median is difficult to make sense of — especially if there was no time to react. At the most likely point of entry, two lanes open for lefthand turns while northbound traffic continues on in the center. There is also a walkway that exits at a narrower point with just two lanes of northbound traffic, but it’s a bit further south.

According to the initial police report, the SUV was traveling at 38 mph and made no attempt to brake. “[Based on video evidence] it’s very clear it would have been difficult to avoid this collision in any kind of mode [autonomous or not] based on how she came from the shadows right into the roadway,” Moir told the San Francisco Chronicle.

Again, the Uber’s lidar should have taken care of that. But, if Herzberg did enter from the median as originally claimed, there is a chance she was obscured by trees until she was already in the road. As the police have not provided the on-board video from the vehicle or established a clear sequence of events, it’s hard to understand exactly what happened.

Moir said the incident occurred within 100 yards of a crosswalk, which would place Herzberg’s most-likely position in the bike line. But the probable entrance point from the median isn’t much further back, so it remains a possibility. We’re hoping the police release the video footage soon. While nobody is excited to see a tragic death caught on camera, it’s the best way to make sense of what happened. We’ve reached out to the Tempe Police Department for clarification and will update accordingly.

Meanwhile, the National Transportation Safety Board has launched its own investigation into the incident. The team consists of Investigator-in-Charge Jennifer Morrison and three additional team members who will examine vehicle factors, human performance, and electronic recorders.

UPDATE: Sgt. Ronald Elcock of the Tempe Police Department confirmed for us that the Herzberg entered traffic from the west side of the road (via the median) and the accident occurred in the northbound lanes. The investigation is ongoing. 

[Images: Uber; Google]

Get the latest TTAC e-Newsletter!

Recommended

69 Comments on “Unpacking the Autonomous Uber Fatality as Details Emerge [Updated]...”


  • avatar
    Sub-600

    Nothing is going to stop this technology from quickly being implemented, there’s simply too much money to be made. Lots of people will die before these vehicles are perfected. It’s a “greater good” situation that nobody will be allowed to vote on. Look both ways before crossing, people.

    • 0 avatar
      Matt Posky

      We’ll see what happens with AV Start but there are only a handful of senators trying to keep it form passing as is. If people are worried, now might be a good time to write a letter.

    • 0 avatar

      “Lots of people will die before these vehicles are perfected.”

      Yes, but lots MORE will die in the scenario where we don’t have the technology and humans keep on piloting their own cars.

      • 0 avatar
        Sub-600

        It just seems intrinsically wrong to make unwitting test subjects out of the citizenry.

        • 0 avatar
          twotone

          Tell that to the pharmaceutical and software industries as well as google, facebook and US electoral system.

          It just seems intrinsically wrong to make unwitting test subjects out of the citizenry.

      • 0 avatar
        thegamper

        “Yes, but lots MORE will die in the scenario where we don’t have the technology and humans keep on piloting their own cars”

        True. But for some reason it is easier to accept a death due to a chance encounter with an idiot, unforeseen event, road hazard, etc than it is being killed because a LIDAR cannot tell the difference between a stationary pedestrian on the side of the road waiting to cross and a mailbox. Like it or not, people will be up in arms when AV’s kill people in scenarios where a human pilot would have easily recognized the situation and avoided a death…..and rightfully so perhaps.

      • 0 avatar
        jalop1991

        “Yes, but lots MORE will die in the scenario where we don’t have the technology and humans keep on texting from the driver’s seat in their own cars.”

        FTFY.

      • 0 avatar
        stuki

        Which would only begin to approach apples-to-apples, if you equipped human driven cars with a similar array of lidar and other awareness enhancers.At a minimum. And even then, the argument is too simplistic.

        IF this car’s multitude of eyes had sightlines to a pedestrian, yet didn’t brake at all until impact, there is something seriously wrong with this machine. Someone has to figure out exactly what. The fact that some/many/most people suck at math, doesn’t excuse a calculator from failing to perform basic multiplication.

        It’s acceptable for humans to behave a bit erratically, as there really isn’t any other option; AND because humans tend to behave erratically in a manner that other humans have learned/evolved to respond reasonably to. Machines don’t follow the latter, hence need to be relied on to behave predictably/non-erratically. Even if one could construct a Robocop that, over a large enough number of instances, could be “demonstrated” to have shot fewer civilians than what human cops currently do, robots running around shooting people at seeming random, are still not an acceptable substitute for their meatsack colleagues.

        • 0 avatar
          Vulpine

          “IF this car’s multitude of eyes had sightlines to a pedestrian, yet didn’t brake at all until impact, there is something seriously wrong with this machine.”

          —- As you point out, Stuki, the question is, “IF”. But you have to remember that these systems in general have been written by humans; humans don’t have the ability to think of EVERY conceivable circumstance ahead of time. While we try to be pro-active, most of us are reactive to events. We have to be. If we tried to anticipate every possible thing that could go wrong in a given circumstance, we’d be afraid to live. And to be quite blunt, I’ve seen far too many people on the roads who are ‘afraid to drive’, going sometimes well below the speed limits and making maneuvers, often at the worst possible time, to prepare for a turn that’s still a mile away. They tend to be the cause of many accidents and are usually the ones who get hit because they aren’t flowing WITH the traffic but rather inhibiting traffic. And no, these fearful drivers aren’t necessarily the elderly, though admittedly many will be.

          Here’s the thing: People need their mobility today more than any time in our history. Many don’t (or won’t) have the advantage of the Internet to have needed supplies and materials brought to them, so they have to go get them. Traffic today is such that someone born almost 100 years ago can’t cope with the levels of congestion we experience today even on wide-open highways. Homes that were once in near-rural environments are now in tighter-than-suburban conditions where sometimes just getting out of your driveway is a challenge. For these people, a self-driving car is a Godsend. Let the car handle the traffic so all you have to worry about is getting what you need and getting it back home.

          And yes, there are those who refuse to use the internet and apparently have no curiosity towards how it can benefit them.

    • 0 avatar
      slavuta

      But we can fill the streets, and protest until it is banned

      • 0 avatar
        FreedMike

        Or we can just refuse to buy these things.

      • 0 avatar
        ajla

        I have no problem with AVs as long as I’m never banned from manual driving on public roads and the situation surrounding accident liability is better defined.

        • 0 avatar
          FreedMike

          I have no problem with autonomous tech for applications like long Interstate slogs.

          (Long haul truckers might feel differently, though.)

          But in everyday traffic, with unpredictable variables like other cars and pedestrians, I see no evidence that this tech is ready for widespread use.

        • 0 avatar
          Lorenzo

          You just mentioned the dreaded word: “liability”. Lawyers use that for their insurance company clients. Lawyers and insurance companies will decide the future of AVs in this, and future accidents. This fatality accident will begin the process, setting the precedents that will decide whether AVs live or die.

          • 0 avatar
            stuki

            Lawyers, and other coercion beneficiaries, looks increasingly set to decide the future of everything in America. Right up until their preying on anyone productive, reduces the incentive to produce to the point where they become overruled by islamists, who haven’t yet fallen so thoroughly for the arbitrary-rule-by-lawyer-is-a-OK, as long as its labelled “rule of law.”

    • 0 avatar
      toxicroach

      You say that like 4000 people don’t die like this every year from normal drivers. Or the 59000 injuries that happen.

  • avatar
    IBx1

    I’m curious to see this pedestrian walkway that says “do not cross.” Is it as blunderous as I am imagining it to be?

    Besides that, how many people could have been driving at night that the pedestrian couldn’t have waited for the car to pass before entering the roadway outside of a crosswalk and during a green light?

    This picture makes the road look a lot larger than I had imagined, and a person paying attention would have been watching the pedestrian walking in the median knowing they intend to cross to one side of the road or another.

    • 0 avatar
      Matt Posky

      The median walkway is VERY strange. It appears to exist only as a way for foot traffic to make it from one side of the road to the other. It does not extend to the crosswalk and empties out directly into the the street, yet has several signs that indicate it is not intended for pedestrians.

      Here’s a link to the area: http://bit.ly/2IE8x7x

      • 0 avatar
        IBx1

        Matt, thanks for the link! That’s definitely a feature that could only be the product of government work. Looks too small to be an official-use u-turn spot, and the curb doesn’t dip to allow it either. Without getting distracted by what is the intended purpose of that thing, it’s clearly marked that pedestrians need to use the crosswalk just up ahead. The pedestrian must have been running straight across the whole road with her bike if the car in the right lane didn’t see her in time.

        • 0 avatar

          As IBx1 says, if I’m understanding what I’m seeing at the link Matt provided, the individual had to move across the southbound lane, across the northbound (in front of the Uber) and nearly make it to the sidewalk on the east side of the road and then be struck by the Uber on the passenger side of the vehicle. That leaves me with the question of how was she NOT seen after crossing nearly the whole of N. Mill Ave.? (If she actually entered the roadway from the West as stated.)

          It makes much more sense if she had been on the East (right hand side) and attempted to cross – or get into the bike lane – not seeing, or hearing, the presence of the Uber,thus being struck by the Uber immediately, resulting in the damage to the right hand side of the Uber. Notice the bike lane AND right turn lane are a combined affair – not entirely separate lanes. If she intended to get into the bike lane, she would have had to move to the left side of that combined right turn/bike lane. Something seems wacky – or I’m completely turned around on my directions. I did check my “bearings” on directions at least 4 times to make sure what I was seeing from a directional point.

          • 0 avatar

            Couldn’t get into edit soon enough. Instead of crossing north and south bound, I meant both north bound lanes. As others have said, IF she crossed from left to right, the vehicle should have seen her unless she was traveling at a velocity at which only the Flash can travel.

          • 0 avatar
            IBx1

            Police released the video up until the impact. Driver was looking down at their phone the whole time but the bike was not visible on the forward-facing camera until it was about 20 feet away. Bike was moving from the median all the way across so I have two questions.

            One, how did the car’s sensors not see a human with a bicycle, a large profile object, moving across the roadway?

            Two, what drugs were in the pedestrian that would cause them to walk a bike slowly across a wide road when the headlights are clearly visible?

            I’ve nearly run 4 bicycles over with my truck over 3 years because they’re riding at night in the road lanes without lights of any sort. Not on the shoulder, not on the sidewalk, and not any faster than 10mph. You really can’t see them until you’re about to make them a human-metal-hybrid hamburger.

          • 0 avatar
            Vulpine

            After viewing the video myself, I’m forced to agree; clearly none of the sensors saw the target in time yet Lidar was arguably the best possible sensor for this specific purpose. (I say arguably as I never agreed and still feel it’s far, far too slow to enable safe driving on its own. Now it’s proving too slow to even support other systems.) Clearly it failed. Now the question is why.

            As for the camera system, said cameras need some form of low-light capability, be it infrared or ultraviolet. This video just emphasized how poorly modern headlights are, especially when trying to view a dark object such as this pedestrian in dark clothing. The first thing visible is the near-white bag hanging on the bike itself.

            As for the distracted driver thing… when I first viewed the video the driver looked more asleep than paying attention to a phone. Didn’t see any evidence of a hand-held device but clearly saw the panic in driver’s eyes when she realized what was about to happen.

          • 0 avatar

            Despite the haircut, that’s a man there.

          • 0 avatar
            Vulpine

            Hard to believe… but you’re right.

          • 0 avatar
            Vulpine

            Then again… maybe not. Other articles are claiming that Rafaela is a woman.

            https://www.cnet.com/news/was-ubers-driverless-car-crash-avoidable-some-experts-say-the-self-driving-car-should-have-braked/?ftag=CAD1acfa04&bhid=20851701915122620595993041636842

          • 0 avatar

            YIKES

      • 0 avatar
        Steve65

        It looks to me like those “walkways” were originally traffic crossovers so they could move vehicles to the other side of the median during construction operations. Why they then brick paved them instead of landscaping is a mystery.

  • avatar

    In the videos of the scene I’m noticing a sprinkler system is running right where the Uber stopped. Makes me wonder if the lack of light and water particles in the air interfered with some sensors.

  • avatar
    sportyaccordy

    The autonomous alarmists collectively jizzed.

    This looks like an accident that might have been unavoidable for anybody or anything. I don’t want to see her death but I hope video can clear things up.

    • 0 avatar
      mcs

      Well, how about this video:

      youtu.be/_CdJ4oae8f4

      The Uber should have detected the bicycle. There’s no excuse. Those sensors can see in the dark better than any human. Like the video above shows, their system has issues.

    • 0 avatar
      Cactuar

      Are you implying that people who question the need or development of autonomous tech enjoy seeing people getting hurt or killed?

      • 0 avatar
        sportyaccordy

        No, what I’m saying is that people who write off autonomous tech with gloom and doom enjoy seeing it stumble and fail. “SEE? SEE? IT KILLS PEOPLE!!” So do human drivers, at much higher incidence rates.

        • 0 avatar
          mcs

          I’m an advocate for AVs, I’m just not happy with some of the current implementations. I’m also concerned that some companies are pushing the technology out before it’s ready. It could taint the market and I don’t want to see that happen.

          • 0 avatar
            sportyaccordy

            I agree that the excitement and proposed roll out dates are premature, but at the same time this tech has to be tested in the real world before going commercial. It’s a tough balance.

            Thankfully, unlike many humans I’m sure Uber will learn from this and do better going forward (if they start this program back up).

        • 0 avatar
          ajla

          *Unsafe* human drivers are dangerous. People drunk, tired, road-raging, texting, etc.

          However, after this incident I’m not convinced that current autonomous technology is any safer than an able-bodied and attentive human driver.

          • 0 avatar
            FreedMike

            Indeed, looking at that stretch of road, the *only* way this driver’s story makes sense if the bicyclist just decided to jump out into the way of a car she should definitely have seen coming. Somehow I have a feeling that’s not the story, unless she was intoxicated, or suicidal.

            It’ll be interesting to see the video.

          • 0 avatar
            Spike_in_Brisbane

            Will drunk, tired, unlicensed, uninsured, criminal or drugged drivers be the ones buying autonomous vehicles?

          • 0 avatar
            Vulpine

            In some cases, yes.

  • avatar
    FreedMike

    My conclusion? Unless the driver was completely asleep at the wheel, he should have had plenty of time to see someone stepping out into the roadway, based on a) the speed of 38 mph indicated, and b) the fact that the road is flat, wide and fairly straight.

    That, or the cyclist just decided to jump out in front of a moving car. I guess that *could* be the case, but it seems less likely to me. We need the video.

    • 0 avatar

      The convicted felon driver might have had his attention elsewhere though!

      • 0 avatar
        FreedMike

        Unless it turns out that this woman just decided to play Human Frogger, and willy-nilly jumped out into traffic despite being able to see a car coming at her at 38 mph, I have a feeling this particular accident was eminently avoidable by either the driver or the car’s lidar systems.

      • 0 avatar
        mcs

        The Uber safety driver in this video wasn’t paying attention either. It would be interesting to check the phone records of the safety driver.

        youtu.be/_CdJ4oae8f4

  • avatar
    Vulpine

    While data is necessarily still very vague and spotty at the moment, a report read less than an hour ago clearly states she stepped off of the curb, into the car’s path. If so, the car, like any driver, would have assumed she would stay on the sidewalk until after the car passed (did the car have any kind of marker lights on? Daylight running lamps? Headlamps?) At least at the moment, the most the car could have done was MAYBE attempt to swerve away from her, assuming it detected her movement at all.

    What this does prove is that the LIDAR is not the panacea some people thought it would be for AV sensing.

    • 0 avatar
      mcs

      The bikes front wheel would have come off the curb first. I’m assuming if she was pushing a bike, she would have been midpoint more or less on the bike. Also, if it was loaded with bags, she wouldn’t have been moving fast.

      • 0 avatar
        Vulpine

        Have you looked at the photos of the bike and vehicle? The strike was on the right front of the car and the bike’s frame appears bent just forward of the pedals. I have no idea which side of the bike she was on when she stepped off the curb. Were she holding the handlebars (likely) her hip would have been just forward of center on the bike itself as she walked it. I haven’t seen any images of where she landed but the force of the hit appears to have thrown the bike itself to the far edge of the roadside sidewalk. She probably landed right at the curb.

        • 0 avatar
          mcs

          Here’s a NY Times diagram of the accident. Although, until we get more information, I’m still not convinced one way or another as to which curb she stepped off of. If it really was from the left/west, the vehicle should have reacted to her. It looks like she would have crossed three lanes. Even if it was from the right, the vehicle should have reacted. Humans move very slowly compared to computer time. A second or two is forever. They should have been pulling camera frames at least 60 to 90 frames per second. After 5 frames the vehicle should have been able to figure out she was going to cross its path.

          If she really stepped off the left curb, there is no excuse. Plenty of time to react there.

          Something odd about this whole thing anyway.

          https://www.nytimes.com/interactive/2018/03/20/us/self-driving-uber-pedestrian-killed.html

  • avatar
    JMII

    Of course the human driver wasn’t paying attention, he was in a computer driven car. Also 38 mph means you cover 55 feet per second. Sorry but even an F1 driver would have likely at least clipped her given the conditions – remember it was dark.

    Checking the Google Street View (posted by Matt Posky above) I see several trees, a large bush, a small hill and a sign off to the right. Just before the sign is a clear pathway and small picnic area. Back up the view to just slightly after exiting from under the overpass. You can’t see anything behind the hill/bush where the turn lane starts. Anyone coming out onto the road from that point would have been hit for sure, sensors or not, its a blind corner.

    If she came from left (the median) then there is no hill but still some smaller trees. In that case there was a small window to spot her.

    • 0 avatar
      Vulpine

      Remember, the LIDAR is on the roof of the vehicle, it should have seen her before the driver could. However, she also made a very unexpected maneuver and stepped out onto the street from the curb (right), according to the latest report I read.

  • avatar
    Testacles Megalos

    If we’re going to replace (supposedly) thinking humans with computer drivers, then the liability for car/non-car crashes must lie by default with the car, and a human liability-bearer prospectively identified.

    Alternatively, legal requirement that all autonomous cars have 360 video recording at all times – memory is cheap – hooked into the pilot software/memory. Since there is no involved human, a case like this has no reliable witness. We deserve video and driving data so that the accident can be reconstructed and true blame assigned.

  • avatar
    DC Bruce

    The problem with all of these systems is that they are reactive. However, the best human drivers recall their training to be “defensive” drivers. Being defensive means both being prepared for unexpected situations (i.e. when someone doesn’t follow the rules) and being predictive about what someone is going to do. So, for example, a driver seeing someone in a median strip, pushing a bicycle, might predict — and therefore be prepared for — that person to step into the street thinking that there’s enough time to safely cross. Or, driving through a residential neighborhood, a driver might see two kids playing ball in their front yard and allow for the possibility that the ball might go into the street, with the kid running after it.

    I don’t think computers and algorithms are at all good at this kind of activity. And we’ve already seen the “if its not moving, I don’t have to worry about it” problem from self-driving cars. If it’s a truck parked on the side of the road, that might be o.k., but if the truck has flashing red lights on top and is red itself, it might be an ambulance that’s about to move. Does the AI know that?

    Similarly, I think the presence of a human “back-up pilot” is falsely reassuring. With nothing to do, its hard for the best of us to be one our toes, ready to compensate for the limitations of the AI. Far more likely, even if not doing prohibited things like fiddling with a smartphone that the back-up driver is listening to music, looking around at the scenery, daydreaming or talking to the passenger in the car. Because that’s what humans do when their attention is not engaged with the task at hand.

    • 0 avatar
      mcs

      DC Bruce: Does the AI know that?

      While I can’t say for sure about the current crop of AV systems, at the bleeding edge we certainly can tell. I even have text recognition built-in to give even more clues. To predict what a person is going to do, I can get data that tells me what direction they are looking (head angles) and what their body mechanics are doing. I also want to analyze front wheel turning motions at some point. It can tell the difference between a human adult and a child. Still has some trouble figuring out gender of people. It can even take a good guess at the breed of a dog. It has a huge vocabulary of objects. Yeah, you can talk to it like HAL. AI is much further along than most people think.

  • avatar
    SCE to AUX

    Just last week, we were treated to video of happy patrons of driverless taxis.

    My question at that time was, “Who owns the liability?”:
    https://www.thetruthaboutcars.com/2018/03/waymo-exhibit/#comment-9486930

    The answer in this case will be Uber, and I hope it’s a 9-figure settlement.

    • 0 avatar
      Vulpine

      I wouldn’t make assumptions if I were you. Early reports absolved the vehicle of fault. The outstanding question is whether the lady was or if she simply never saw the car coming.

      • 0 avatar
        SCE to AUX

        “Absolving the vehicle of fault” means the social contract for AVs needs to be clarified.

        A hands-free vehicle fitted with the latest situational awareness technology *cannot* be allowed get away with involuntary manslaughter.

        An even worse scenario would be this: an AV strikes a human instead of a barrier, or chooses to strike human #2 instead of human #1, all in the interest of protecting its occupants.

        At some point the mfr needs to own the liability, and I think we’re already there.

        • 0 avatar
          Vulpine

          Again you’re making assumptions on insufficient data. I clearly stated, “Early reports said…” Nobody, not you, not me nor anyone else here has all the data. Let’s wait and see rather than jumping to conclusions. Until then, you’re more likely to embarrass yourself than offer anything constructive.

        • 0 avatar
          ktm

          Sorry, but “autonomous” driving is a misnomer. The human occupant is the real “driver” and must pay attention at all times. Period. That is how the liability will work itself out. If the system fails, the human must take control. Now, if the system will not let the human take control, that is another matter.

          My concern is maintenance. I have yet to hear anyone talk about this issue and I raise it every time I can. We have people who do not change their oil for 80k miles. Now we will have vehicles with complex systems that need to be maintained by said people.

  • avatar
    ixim

    I’m not a computer expert; I merely own several of them and use them all the time. A/V’s are coming; there’s a huge amount of $$, engineering muscle and, yes, political clout behind their arrival. And when that happens, they will mix with all the other users of the roads and streets. Those humans – drivers, riders, walkers – need to know how A/V’s behave. Will they all be programmed with the same parameters/protocols? Can we depend on all of them having the latest software updates all the time [right!]? Undoubtedly, A/V’s will injure people and damage property. The owners of these vehicles MUST be held liable. Will they, or will we see their victims being blamed? I get A/V’s utility for those who cannot drive. Of the billions of miles humans drive every year, most do not result in any harm. Will A/V’s really do better?

  • avatar
    TheEndlessEnigma

    There will be, and is, a rush to absolve the automated system from any liability. Regardless of the circumstances here the automated system WILL NEVER be found to be at fault. Too much money and political bull$hit is invested in this who automated system scheme.

    • 0 avatar
      Vulpine

      Conspiracy theory. One thing about the modern systems is that they tend to record their circumstances on a regular basis… almost like a built-in dash cam. Visual proof of an incident is far more reliable than hearsay and even eyewitness accounts, because the eyewitness tends to either see the event after the fact or not see the actual cause of the incident. A camera is impartial and therefore more trustworthy.

  • avatar
    roverv8i

    I feel like the core issues are being missed in this discussion. Regardless of the flaws in Human thinking we have a superior ability to navigate the world compared to computers and current AI. If we want to progress to true autonomous vehicles we need to come to terms with the fact that Humans will have to give up certain perceived “rights” on the roadway. While it may be true that the driver and/or autonomous car should have been able to avoid the collision, it does not change the fact that the woman was jaywalking as there was a crosswalk nearby. If she had proceeded to the crosswalk who knows if she would have been hit but at least she would have been where she was supposed to be.

    Ideally we would redesign the roads to help make it easier for autonomous vehicles to navigate them. That would not just require changes to the physical infrastructure but also human behavior around said roads. Were we to do this I believe current tech is more than sufficient to handle autonomous duties. For autonomous vehicles to work in the current infrastructure a big leap is required in the AI. Yes, there are issues with current systems being able to see certain things but the core issue is that they do not know what to think when they can’t see certain things or are presented with an new situation.

    We will ether have to dumb done the infrastructure or make AI more intelligent. If you think about this you will realize that the human element of driving can not be replicated by simple 0’s and 1’s. I will leave it to your imagination to realize the potential implications of this.

    • 0 avatar
      Vulpine

      @rover: Ever watch much anime? Ever notice how they discourage jaywalkers? That’s right, waist-high barriers, like many of our highway guardrails, between street and sidewalk (actually, right on the curb) except for intersections and legal crosswalks. Now, imagine the squawking in the US if such were made mandatory here.

  • avatar
    Garrett

    Why stop at vehicles?

    Autonomous medical treatment. When it messes up, who gets blamed?

    Autonomous financial management. When people loose their retirement savings, who gets blamed?

    Autonomous firearms for hunting or self defense. When those friendly Mormon missionaries get gunned down, who gets blamed?

    Autonomous food prep. When people get food poisoning, who gets blamed?

    Unless we can send the owner, attendant person, and everyone in the software and hardware development train to prison, and I’m talking real prison where getting shived or raped is a possibility, we have no business deploying this stuff.

  • avatar
    lon888

    Read about the fates of the early explorers and their shipwrecks. Read about man’s early attempts at powered flight.
    Read about man trying to fly passenger jets i.e DeHavilland Comet
    Read about man’s early space exploration accidents.
    Moral of the stories – any time a new technology is tried accidents will happen. This is just the first one – deal with it.

  • avatar
    PandaBear

    Firmware engineer here.

    These kinds of accidents will happen when technologies first started, especially during war time because the risk of not acting is worse than the risk of your test pilot dying or your solders killed with failed equipment.

    It is unfortunate, and it is not avoidable but hopefully all parties involved do the best they can to minimize the probability of it happening. The “driver” should be at least a professional that collect logs and test the car / software with a “fail safe” attitude that makes a failed software do a minimum amount of damage. It is better to have a computer shut down the car or do a false emergency brake than to run over a pedestrian, or cause the car behind you to rear end you than you rear end the car in front of you.

    Remember: there is no bug free software, and there is no “perfect” human driver. Hopefully the software is safer than the human driver, and is tested as much as possible.

  • avatar
    ixim

    Three video screenshots show the victim walking her bike into the road from the left clearly illuminated by the headlights. Although the A/V didn’t apply the brakes, it’s not clear if the human “driver” could have stopped in time either. The victim nearly made it across as it’s the right front fender that was damaged. It’s still not OK to hit anyone/anything with a motor vehicle; accidents can and will happen. I see no proof that A/Vs will be any better or worse than we humans in that regard.


Back to TopLeave a Reply

You must be logged in to post a comment.

Recent Comments

  • kosmo: “How’s that Ranger Raptor coming, Ford? Oh, it isn’t? I see. Thanks for the mobility scooter,...
  • dividebytube: When I’m down south I’m taken aback by the number of decent looking old trucks and even G...
  • redapple: RED…. Great catch. Love it.
  • teddyc73: What an ugly rear end.
  • FreedMike: Have you ever heard a Ghibli’s exhaust? I’d say that’s one big selling point. Otherwise,...

New Car Research

Get a Free Dealer Quote

Staff

  • Contributors

  • Timothy Cain, Canada
  • Matthew Guy, Canada
  • Ronnie Schreiber, United States
  • Bozi Tatarevic, United States
  • Chris Tonn, United States
  • Corey Lewis, United States
  • Mark Baruth, United States
  • Moderators

  • Adam Tonge, United States
  • Corey Lewis, United States