"The victim did not come out of nowhere. She's moving on a dark road, but it's an open road, so Lidar (laser) and radar should have detected and classified her" as a human, said Bryant Walker Smith, a University of South Carolina law professor who studies autonomous vehicles.
Smith said the video may not show the complete picture, but "this is strongly suggestive of multiple failures of Uber and its system, its automated system, and its safety driver."
I, For One, Welcome Our New Self-Driving Overlords
Categories:
29 Comments
Leave a comment
Best Canadian Blog
2004,
2005,
2006,
2007
About Kate
Why this blog?
Until this moment
I have been forced
to listen while media
and politicians alike
have told me
"what Canadians think".
In all that time they
never once asked.
This is just the voice of an ordinary Canadian yelling back at the radio - "You don't speak for me."
homepageemail Kate
(goes to a private mailserver in Europe) I can't answer or use every tip, but all are appreciated!
Support SDA
I am not a registered charity. I cannot issue tax receipts.

Want lies?
Hire a regular consultant.
Want truth?
Hire an asshole.
Polar Bears: Outstanding
Survivors of Climate Change.
Dow Jones
What They Say About SDA
"Smalldeadanimals doesn't speak for the people of Saskatchewan" Former Sask Premier Lorne Calvert"I got so much traffic after your post my web host asked me to buy a larger traffic allowance." Dr.Ross McKitrick
Holy hell, woman. When you send someone traffic, you send someone TRAFFIC. My hosting provider thought I was being DDoSed. - Sean McCormick
"The New York Times link to me yesterday [...] generated one-fifth of the traffic I normally get from a link from Small Dead Animals." Kathy Shaidle
"Thank you for your link. A wave of your Canadian readers came to my blog! Really impressive." Juan Giner - INNOVATION International Media Consulting Group
I got links from the Weekly Standard, Hot Air and Instapundit yesterday - but SDA was running at least equal to those in visitors clicking through to my blog. Jeff Dobbs
"You may be a nasty right winger, but you're not nasty all the time!" Warren Kinsella
"Go back to collecting your welfare livelihood."Michael E. Zilkowsky
Intelliweather
Seismic Map
Comments Policy
Best Of SDA
Hide The Decline
The Bottle Genie
(ClimateGate links)
You Might Be A Liberal
Uncrossing The Line
Bob Fife: Knuckledragger
A Modest Proposal (NP)
Settled Science Series
Y2Kyoto Series
SDA: Reader Occupation Survey
Brett Lamb Sheltered Workshop
Flakes On A Plane
All Your Weather Are Belong To Us
Song Of The Sled
The Raise A Flag Debacle
(Now on Youtube!)
(.mwv Video)
Abuse Ruins Life Of Girl
Trudeaupiate
Kleptocrat Jeans
Child Labour
I Concede
Small Dead Feminist
Protein Hoser: THK Interview
The Werewolf Extinction
Dear Laura (VRWC)
We Wait
Blogging The Oscars
Jackson Converts To Islam
Just Shut The HELL Up
Manipulating Condi
Gay Equality Rights
Blogroll
News Aggregators, Resources
Drudge Report
Bourque (Canada)
Memri (Middle East)
Newsmax
Military News Spotlight
Watching America
Int. Free Press Society
Newsbeat1
Rawlco local news
Dates in History
Newseum
Oilprice.com
My Westman
Favorites
Instapundit
Weekly Standard
Outside The Beltway
Day By Day
Hugh Hewitt
Mark Steyn
Belmont Club
Powerline
Den Beste (archived)
American Thinker
Victor Hanson
Michelle Malkin
Michael Yon (Iraq Imbed)
Tim Blair (Oz)
Protein Wisdom
Captain Capitalism
Kathy Shaidle
David Warren
Damian Penny
Publius
Newsosaur
Edward Michael George
Long War Journal
Eric Anderson
Inconvenient Science
Climategate 2.0:
The emails unredacted
Search the database
Climate Audit
Icecap
Anthony Watts
Climate Debate
HK Climate
Climate Depot
Energy Matters
Yanks, mostly
Professor Bainbridge
Stephen Green
Wizbang
Daniel Drezner
Dean Esmay
Right Wing News
Patterico
I Could Be Wrong
Maggies Farm
Bill Roggio
Musing Minds
Pajamas Media
Newsbusters
Blackfive
Day By Day
Cox And Forkum (archives)
The Last Tradition
Canadian, eh?
Torches And Pitchforks
Bow. James Bow
Ghost Of A Flea
The Black Rod
Proud To Be Canadian
Quotulatiousness
Uncle Meat
Halls of Macadamia
Full Comment (NP)
Steynian
Blazing Cat Fur
Inspiringyoutothink
Prince Arthur Herald
March 2018
Recent Comments
- TheTooner: https://jalopnik.com/uber-has-no-damn-business-testing-self-driving-cars-1823982487 read more
- Kenji: Exactly right. Humans can make “connections” ... like ... (deer read more
- historybuff: the jaywalking in my town is worse than anywhere Ive read more
- Boots: Or deer and other wildlife Steve. Anyone that's driven country read more
- Steve from Rockwood: Another tough issue for auto-driving cars is when someone (or read more
- TimR: Really good points. The other aspect of situational awareness is read more
- Zon: Negative. Autopilot can be engaged as little as a few read more
- Boots: That's what I take out of this also Richard. The read more
- The Phantom: Tempe AZ, the speed limit on a road that size read more
- coldinsaskabush: Did they check to see if the safety driver had read more







"Safety Driver". An ex-con and a trans freak show. Yeah no.
Changing the face of liability matters. Insurance companies are overjoyed.
I can't imagine more boring job than being a safety driver!
But we should stop focusing on "stupid" accidents only, we should to consider broader statistics as well (we should get some by now, so where are they?).
And here is my question: Should we accept technology which could reduce road fatalities by 70%? By 90%? By 99%?
Just a thought.
If it was me driving, I'd have hit her.
This whole episode is what I'm starting to refer to as "Picasso News". No matter what angle I look at it from, nothing makes much sense.
saw the video, it is not just uber at fault. When I walk on a dark road at night, I take care. It is difficult NOT to notice that a car is coming, the headlights usually are a clue. It was a straight road, at least for hundred meters near the scene. The woman did not even appear to notice that a car was coming, just where exactly did she think she was??
I don't like the idea of automated cars, too many risks go along with it. However, pedestrians crossing roads at night, well - what can I say - you've got to know what you are doing as well. The law of man is fickle, the laws of physics - not so much.
Here's the link to the Le Devoir article https://www.ledevoir.com/societe/environnement/523034/deux-importants-projets-industriels-echappent-au-nouveau-test-climat
So all these commercials being shown recently on TV showing drivers narrowly avoiding an accident because the vehicle's sensors picked up an object in its path are B.S. Good to know.
I’ve missed hitting plenty of animals at night that were closer than the poor woman.
People should be asking why this is being sold as unavoidable when it clearly was avoidable
It was unavoidable. The car was controlled by a computer program, i.e. by computer programmers who weren't there, who had already made all the decisions about everything they could imagine about a driving situation they wouldn't be in because it hadn't happened yet. And the backup to this was a human "driver" who wasn't driving and wouldn't be taking over until, and if, he/she noticed the automated system which was better at detecting and fast at dealing with emergent hazard situations, hadn't detected one and wasn't dealing with it.
That combination cannot avoid all collisions. This stuff is going to happen with these cars, it's unavoidable. And no, we can't avoid it by not having self-driving cars. We've invested a couple of generations in becoming a society collectively too stupid, too lazy, and too self-absorbed to reject this technology. It will take a lot of stuff much worse, happening to almost every individual, to get people to give up the "smart phone" culture.
How can they justify using a video shot at night as proof the collision was unavoidable. Anyone who has ever attempted to shoot video in nighttime bright/low light conditions notices that the camera aperture does a very poor job of capturing and displaying objects that are illuminated less then others, and displays that part of the image as black, even though you can clearly see these objects with the human eye. A camera video recording in nighttime road conditions clearly cannot reproduce what millions of years of evolution has done to make the human eye the best camera ever made, and should never be used as definitive evidence of anything. I am surprised that this video is being used this way and that the pedestrian was invisible to drivers on the road. From what I can see of the video, it is clearly a case of distracted driving caused by all the fancy dashboard autonomous displays.
The Three Laws of Robotics (often shortened to The Three Laws or known as Asimov's Laws) are a set of rules devised by the science fiction author Isaac Asimov. The rules were introduced in his 1942 short story "Runaround" (included in the 1950 collection I, Robot), although they had been foreshadowed in a few earlier stories. The Three Laws, quoted as being from the "Handbook of Robotics, 56th Edition, 2058 A.D.", are:
1.A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2.A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
3.A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws.[1]
Exactly. This is being SOLD as unavoidable. Nonsense. The human eye SEES much more than the video cam is revealing ... starting with movement ... motion ... even of objects in the dark. This JUST happened to me when in total darkness my eyes detected ... motion ... on the shoulder about half mile ahead of me. I couldn’t make out what it was ... but my HUMAN eyes and brain told me to slow down ... and it ended up being a human walking along the shoulder.
IF this “fantastic technology” is being sold as superior to human ability to drive safely ... then that is a LIE. Human accidents are mostly caused by crap drivers. Inattentive drivers. Drunk drivers. Distracted drivers. Yes ... a self-driving car will be an improvement over THOSE crap drivers. But it will always be inferior to a competent human driver.
The pro-self-driving lobby reminds me of the gun-grabbing, anti-2nd Amendment lobby. Guns that are owned and operated by competent, non-criminal humans are completely safe. Guns operated by irresponsible and criminal humans are dangerous. The anti-gun ... anti-car people are really anti-human. They refuse to live with fallible humans. We humans have improved and evolved automobile technology to make our current autos incredibly reliable, safe, and comfortable. The operation however, should remain in human hands. You will pry the steering wheel from my dead cold rigor mortised hands ... as with my guns (that I actually don’t own yet).
The thing I always wondered about the first law was, "Is death considered an injury, or in the logic-cycling brain of an AI, is the case specifically included, because injury can only happen to the living?"
Does anyone else here see the irony in relation to those TV commercials about existing safety systems?
Instead of back-up A-I correcting a human driver's critical mistake, we have a back-up human failing to correct an A-I driver's critical mistake.
Anyways, we seem to have entered a brave new world where people actually die as a direct consequence of 'computer glitches'. If my last POS HP laptop was any indication of current industry reliability, we will see carnage on our streets.
So, if I use my self-drive car for say, 10 years, will the manufacturer start intentionally degrading the safety or performances modes in an effort to convince me to buy a new one. It's not like certain companies, like apple or microsoft, aren't know to do this?
The unintended as well as the intended glitches, errors, whatever you want to call them are endless. Of course, everyone will want to make money on this but no-one will want the liability of it. Don't worry, I am sure our gov't regulators will step up to ensure it is all safe, perhaps someone should ask CSA to start looking into it. This is the kind of graft and corruption that they live for....
I'm already dissatisfied with the performance of my automatic transmission. I can chose the gear I want better than the transmission. (Let's not have the F-1 conversation, shall we? We don't have the F-1 transmission, we have the 6 speed Ford/Chevy one that can't settle on a gear in the snow.)
There a reason for that. I'm actively driving the car, reading the conditions and planning ahead. The transmission doesn't plan ahead, it -reacts- to conditions. It is always behind the 8 ball, because it doesn't know what's going to come next. I do, that's why I downshift in a corner, or choose 3rd gear in the snow instead of second.
The self-driving car doesn't "know" what's going to happen. It can only react to conditions, it can't plan ahead for them. It is a machine. It can't know things.
This accident is a bad one. Two humans screwed up, the safety driver and the victim, AND the car did not react to the large object in the road in front of it. That's a major fail on the car's part. The thing has radar, it is supposed to be able to see stuff like a large solid object in the road.
The sales job is because this self-driving thing is coming, and it doesn't matter what we say. They're going to replace every human commercial driver with a robot, in the next 10 years, come hell or high water. It will take a violent insurrection to stop it.
And you know, there might actually be a violent insurrection over this. People hate these machines. I've driven behind them, they are the worst combination of little old lady and eco-freak safety nazi you've ever seen. They do not drive like a normal human. They drive like somebody you want to beat up.
You see one in a week, its a curiosity. You see one a day, its that thing you know not to get behind. You have your drive to work screwed up by a flock of these things every day, you're going to start being extremely aggressive with them on the road. Multiply by every driver younger than 60 on the road, that's an exceptionally hostile environment for self-driving cars. People are going to be brake-checking them, crowding them and not letting them merge. I strongly doubt they can handle it.
I see the entire corporate media lining up to smooth all this over far, far too quickly. Every news story I've read on the accident is blaming the biker and repeating the narrative that 'nothing could have been done' to avoid the accident. Well, you can't say that, because you can't relive the moment with a real human actively driving the vehicle.
A couple of points:
I guarantee that bike had reflectors on it that the headlights would have shined on and the driver would have seen. The packs too. There's more to this accident than this highly cut and compressed video. Also, wtf is the point in having a "safety driver" behind the wheel on their phone the whole time. Had she/he/it been watching the road, it probably wouldn't have happened.
Canadian Observer is right. We should have AI warning the human driver and/or his mistake, not the other way around. Human nature mitigates against the other way. If the AI has been doing all the driving, the human backup cannot possibly be as attentive as if he were the only driver. Quite the contrary. He may be playing with his I phone all this time, and not paying attention at all.
Fighter jets have much more complicated controls, and much more elaborate processors. But the pilot, admittedly much better trained and qualified, controls the plane, with the help of information from the processors, not the other way around.
Commercial airplanes have autopilots. But that is only used when the plane is cruising, when it is at high altitude with no other planes around it. I don't believe they are used in a holding pattern, for instance. Yes, there is instrumental landing, but only under conditions when the pilot cannot see (such as heavy fog) and has to rely on the radar where the ground is. The ground crew makes sure that the entire vicinity of the runway is cleared, as well as lighting up the runway as much as possible. I believe any commercial pilot will take over as soon as he can make out the runway.
TheTooner is right in his ironic declaration. The computer can only process information based on its algorithm, which cannot possibly be programmed against all eventualities. Even if the programmers can think of all of them, the onboard computer cannot possibly handle all of that in real time. For example, Humans have defeated computers in chess until recently. The computer is not a strategist. What it can do is look so many moves in advance, for all possible moves. If a human can launch a combination involving more moves than the computer can process, then it will not realize the danger. I believe the recent victories by computers is due to increased processing capability of the super computers used. If it can see as many moves ahead as the best grandmasters, then of course it will win.
I have no objections to AI being used on highways. That is as near the commercial airplanes cruising as possible. The variables are much less on a restricted access highway, and may be within the capability of the computer to take them all into account. The human driver should still be in control, especially in heavy traffic. However, basically using "autopilot" for most of the time in long distance driving will keep the human driver fresh for longer stretches.
I have also no objections to warnings as in a jet fighter, such as collision avoidance, in a city street scenario. But those warnings should be used, just as in a jet fighter, to assist human judgement, not instead of. Using videos of a dark road to excuse the AI is laughable. "Darkness" per se should not affect a self driven car equipped with radar, which generates its own radiation, in a different part of the spectrum. It should be able to differentiate a moving object, in the dark (in the visible spectrum), whatever it is.
Kenji is right in that most accidents are caused by drivers incapacitated in one way or another. The solution to that is remove those drivers as much as possible. Drivers under the influence of any chemical, not just alcohol, for instance. I should be extra careful in states that have legalized marijuana. Tests should be developed to detect those other chemicals in the blood stream. Have a separate test and license designation for limited access highways. Some people just simply do not have the reflex to handle emergencies at high speed. A virtual driving computer that tests those reflexes should be withing the budget of state highway administrations. Set realistic but enforced speed limits. Have the highway police be more vigilant against overaggressive drivers who are accidents waiting to happen. All patrol cars should be equipped with hood cams, and convictions should be open and shut. The penalty should be as least equal to driving under the influence. Do all of that, and the highways will be much safer. They may become so safe that even self driving cars can handle them.
It would be nearly impossible for a radar detection system to classify something as human - even with a super computer. It does not have the resolution. Certainly laser detection has the resolution but it does not "see" very far ahead (I've read about 75 m).
Not sure how fast the car was traveling but I think we need to be reminded just how limited these AI systems really are.
Did they check to see if the safety driver had his pants down around his knees when the accident happened? That is the obvious explanation.
Tempe AZ, the speed limit on a road that size will be ~40-45mph. The traffic will be going 50-55mph. So, pretty quick but not highway speed. The self driving car will be going the limit or a little under, that's why everyone hates them. They slow traffic like an 80 year old Minnesota man in a Cadillac.
It doesn't have to recognize human from large farm animal, it just has to see "big thing in the road!" and miss it. Massive fail there.
That's what I take out of this also Richard. The SAFETY driver. The video from inside the car clearly shows she (or he...or zhe) was more interested in the smart phone than being on the outlook. Isn't that the purpose of a SAFETY driver in these vehicles? When hired are they not trained to NOT be distracted by their devices? Or else what's the point. Good drivers are constantly checking their surroundings. Its called "situational awareness". In this case it might have been noticed that a barely visible cyclist was ahead. Now your paying attention. You notice it's not a Lance Armstrong type cyclist, it's a homeless type cyclist. That means unpredictable. You've learned that from experience. In a split second your brain has you prepared. Decelerate, maybe hover your foot over the brake pedal. Stay focused on the cyclist. None of that happened in this case. If Uber has strict rules for their SAFETY drivers regarding smart device distractions then the driver should be charged with distracted driving. If Uber doesn't then they have some explaining to do.
Negative.
Autopilot can be engaged as little as a few hundred feet off the ground. This is an aid to the pilot during the busiest times in the cockpit, take off and landing.. And yes, the autopilot is used in a holding pattern, holding pattern areas are marked on the charts/databases and the automation can fly them, no problem. Flying seems easy to the uninitiated, but during take off and landing the pilots are BUSY! The old adage rings true: Aviate, Navigate, Comunicate! It would be like driving your car, while the whole time giving a running commentary to a person controlling traffic, all the while being disconnected from the roadway and travelling in 3 dimensions! Aircraft also have separate and redundant systems like TCAS; Traffic Collision Avoidance System, which will warn of an impending collision between 2 aircraft, which that 'talk' to each other, give the pilots advanced warning, if not rectified will give resolution messages (you climb, I'll descend) and if nothing is done, can actually steer both aircraft away from each other. We have what is called RVSM airspace, where the vertical separation is between aircraft is only +- 1000 feet, at high altitudes and requires a whole host of more accurate instrumentation to even enter! There is also EGPWS or Enhanced Ground Proximity Warning System which does similar with non moving terrain and obstacles. The radar altimeter tells the plane how high off the ground the plane is under a certain altitude, but the navigation system tells where the plane is, situationally, radar is not used for any accurate ground mapping, in that sense. The most advanced planes have: Auto throttle, auto brake and auto land.. With 2 or sometimes 3 pilots monitoring all this technology and even still, there are crashes!
Really good points. The other aspect of situational awareness is knowledge of the particular road. A regular driver may have driven the same road several times before and have prior knowledge of the pitfalls. I think any of us that drive regularly know of several spots where we automatically perk up and pay more attention, maybe because we have had close calls at the spot before or we know that people do stupid things in that area. It's defensive driving.
Another tough issue for auto-driving cars is when someone (or something) enters the roadway at right angles to traffic (like someone crossing the road). Most of the technology is collision avoidance (what is in front or behind). But pedestrians will enter the roadway quickly from the side.
Or deer and other wildlife Steve. Anyone that's driven country roads know how wildlife dart across roads. And these roads are not lite up at night like city roads so awareness is key. When you see deer in the ditch you change how you drive. You sit up a bit, maybe slow down some... not sure why but I always turn off the radio. I'm sure that technology will be there soon the way things are going but it's not there yet. It's still Work In Progress.
the jaywalking in my town is worse than anywhere Ive ever seen.
these doped out 'walking dead', jeezuz murphy, right out into traffic. oblivious, death wish stuff.
really bad and getting worse on all but the very busiest 6 lane type streets.
cue the darwin awards !!!
Exactly right. Humans can make “connections” ... like ... (deer in the roadside ditch may spook and lurch across the road) and take evasive or precautionary actions AHEAD of an incident. The same goes for when I am approaching flashing emergency vehicle lights that I see up ahead in the distance ... I cut my speed, and may move over a lane out of caution. Will AI make these “connections” that humans routinely make ?
https://jalopnik.com/uber-has-no-damn-business-testing-self-driving-cars-1823982487