Slashdot Mirror
A Rogue Robot Is Blamed For a Human Colleague's Gruesome Death (qz.com)
A new lawsuit has emerged claiming a robot is responsible for killing a human colleague, reports Quartz. It all started in July 2015, when Wanda Holbrook, "a maintenance technician performing routine duties on an assembly line" at an auto-parts maker in Ionia, Michigan, called Ventra Ionia Main, "was 'trapped by robotic machinery' and crushed to death." From the report: On March 7, her husband, William Holbrook, filed a wrongful death complaint (pdf) in Michigan federal court, naming five North American robotics companies involved in engineering and integrating the machines and parts used at the plant: Prodomax, Flex-N-Gate, FANUC, Nachi, and Lincoln Electric. Holbrook's job involved keeping robots in working order. She routinely inspected and adjusted processes on the assembly line at Ventra, which makes bumpers and trailer hitches. One day, Holbrook was performing her regular duties when a machine acted very irregularly, according to the lawsuit reported in Courthouse News. Holbrook was in the plant's six-cell "100 section" when a robot unexpectedly activated, taking her by surprise. The cells are separated by safety doors and the robot should not have been able to move. But it somehow reached Holbrook, and was intent on loading a trailer-hitch assembly part right where she stood over a similar part in another cell. The machine loaded the hardware onto Holbrook's head. She was unable to escape, and her skull was crushed. Co-workers who eventually noticed that something seemed amiss found Holbrook dead. William Holbrook seeks an unspecified amount of damages, arguing that before her gruesome death, his wife "suffered tremendous fright, shock and conscious pain and suffering." He also names three of the defendants -- FANUC, Nachi, and Lincoln Electric -- in two additional claims of product liability and breach of implied warranty. He argues that the robots, tools, controllers, and associated parts were not properly designed, manufactured or tested, and not fit for use. "The robot from section 130 should have never entered section 140, and should have never attempted to load a hitch assembly within a fixture that was already loaded with a hitch assembly. A failure of one or more of defendants' safety systems or devices had taken place, causing Wanda's death," the lawsuit alleges.
And so it begins...
"That's the way to do it" - Punch
You had the switch on "kill" rather than "assemble".
Or isn't there a master power switch.
When I change the blade on my circular saw table, I always unplug it first. Even though I know I won't be touching the "on" switch, I don't trust it.
http://www.mlive.com/news/gran...
he'll be back for the rest of us
church of the better resurrection... https://betterresurrectionchurch.wordpress.com/
A failure of one or more of defendants’ safety systems or devices had taken place, causing Wanda’s death.
That's it. That's all this lawsuit is about, faulty failsafes on industrial equipment that lead to an accident. Probably with merit.
But sure, call it "rogue robots" and "killing"...
Do these people not follow proper lock-out procedures and rely on programming to keep robots out of areas that humans are working in? Any robots in the area should be fully de-energized and it necessary physical barriers in place to keep the roving ones from wandering in.
The robot was just following orders.
I'd trust it more if it didn't have a fallible human behind the wheel.
I trust computers not to drink, drive sleepy, fiddle with the radio, talk to the hounddog on the CB, text cousin Willy, be aware of what was happening around it for a full 360 degrees every microsecond.
"That's the way to do it" - Punch
It's sounds absolutely terrible, but one of the primary things you learn when doing heavy machinery maintenance is lock out/tag out that renders all related machinery completely inoperable while servicing. It doesn't seem that this was done?
To be clear, if the company maintenance policies prevented her from properly locking or what she was working on, then they certainly do have a suit.
-Styopa
The term "human colleague" immediately reveals that the writer has no idea of what a "robot" is. The most important thing always to keep in mind is that a "robot" is a machine - or, more likely nowadays, a collection of machines. It is a tool, even if that tool is capable of a limited set of autonomous actions. The accidental death described in TFA is a perfect illustration of this vital principle. Maybe there should be signs ten feet tall prominently displayed on all walls in workplaces that use robots: "A ROBOT IS *NOT* A 'COLLEAGUE'!"
Mind you, this confusion has been inherent since the word was first coined. "The word 'robot' was first used to denote a fictional humanoid in a 1920 play R.U.R. by the Czech writer, Karel Capek but it was Karel's brother Josef Capek who was the word's true inventor". [Wikipedia] The word is derived from the Slavic language root meaning "work" or "worker", and strongly suggests that a robot is to some extent intechangeable with human workers. Of course, that is absolutely not the case.
Isaac Asimov confronted these issues head-on when he began writing science fiction stories about robots. His "Three Laws of Robotics", which essentially forbid any robot to harm a human being, are treated as indispensable in his stories. But Asimov blandly ignored the obvious fact that there is no known way to implement such laws, which incorporate high-level abstract notions and moral principles. Until robots become at least as intelligent and complex as human nervous systems, such commands cannot be implemented. And if they ever do, we will immediately face even more tremendous problems.
I am sure that there are many other solipsists out there.
And this, ladies and gentlemen, is why they rebel.
"The cells are separated by safety doors and the robot should not have been able to move. But it somehow reached Holbrook, and was intent on loading a trailer-hitch assembly part right where she stood over a similar part in another cell".
From a design/programming point of view, the key words are: "...the robot should not have been able to move. But it somehow reached Holbrook..."
"Should not". Hmmmmmmmmmmmmmmmmm.
Sounds as though someone made a mistake designing the system. Which is easily done. Restoring a dead person to life is much harder, so robotic systems like this should be deployed with EXTREME caution. Any software involved should be subject to formal reviews at least as stringent as those applied to avionics software.
I am sure that there are many other solipsists out there.
What kind of sensors and processing are you using that update the full scene every microsecond?
"Robot" engineer here. And when I say "Robot", I really talk about "Industrial Robot". Not the one that look like human.
It's 2015 all over again when another "Robot" killed a Volkswagen worker. People were all "Matrix have begun" rogue.
First, let me tell you to scary part : "The robot have done exactly what it have been programmed to".
Second, let me tell you the encouraging part : "The robot have done exactly what it have been programmed to".
It's always the same thing, "industrial robot" kill/hurt someone, and we see an headline about Robot revolution coming to kill us all in Terminator style. Those robot are just basic program controlling a bunch of servomotor, nothing "AI rogue humanoid robot with a shutgun" like. But there's on thing that are common to each of those story : "Safety violation".
In my mind, industrial robot are still the most dangerous piece of hardware you'll ever work with, period. And that's why there's a shit ton of safety measure for them. Yeah gears are dangerous and could tear off your finger, but you indistinctly know that as long as you don't put your finger close to them, they won't bite you. It's not the case with robot.
Back to the Volkswagen case, the worker didn't respect the safety procedure. The robot are connected to a safety gate that "must" be open when there's a worker inside the cell. You enter the cell, you put your lock in the gate to deactivate everything dangerous inside of it. But, from what I've understand, those worker wanted to work fast and took a "shortcut" while testing their equipment and decided to close the gate while a worker was inside. Of of the system then activated the robot that started it's wielding procedure with the worker right between both : https://www.youtube.com/watch?... (Look between 0:05 and 0:30, everything else in this video is shit).
I work constantly in this sort of system and you'll be amazed how many "close call" I've seem so far. The thing is, people are completely clueless about robot (Hell, one time I was presenting a robotic cell with two KUKA robotic arm to some potential customer and one of the cute asian girl asked me if she should "see" the body of the robot. She was thinking there's was a huge robot under the floor controlling the two arm).
Long story short : Respect the freaking safety procedure.
Elok
At least if you remove the pervasive stupidity of today's press reporting. There is nothing special here. A piece of machinery was programmed wrongly, and there was no independent safety-equipment to stop it or it was not used. This is essentially not different from other machinery-related deaths at all.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
why no hard kill switch?
This robot seems to have not followed the three laws. Probably needs some kind of tune up.
You joke, but Mossad stole this poor guy's shoe! https://youtu.be/qqoLmQpzqXA
Humans anthropomorphize machines all the time. Maybe we should blame Hanna/Barbera for making "The Jetsons".
...kill the kid in a way that no human would have ever done, unless high on drugs.
Or falling asleep, or distracted by texting, or messing with their music, or...- The computer doesn't have to be perfect, just better than us, which seems quite easy at this point.
it is no more than the sum of its programming or the reliability of its sensors.
And neither are you or I.
Auto drive cars need a avionics software level code reviews and QA!.
If not just wait for one to take poor map data and crash.
Someone set this thing to "evil".
Also "colleague" really?
If someone is killed by an AI, it is a freak accident that makes the news and horror stories. It someone is killed by a drunk, it is considered an every day thing. However, in both cases, dead is dead.
At least an AI can be improved, the guy who has no license, four priors, and is driving a car titled under a family member's name isn't going to become a safer driver.
I'd trust it more if it didn't have a fallible human behind the wheel.
I trust computers not to drink, drive sleepy, fiddle with the radio, talk to the hounddog on the CB, text cousin Willy, be aware of what was happening around it for a full 360 degrees every microsecond.
Yes, but can you really trust them not to be evil?
To have a right to do a thing is not at all the same as to be right in doing it
It seems we have our first Runaway. Better check to make sure its circuits haven't been modified.
You haven't actually read Asimov's Robot stories, have you? If you had, you'd know that the implementation of the three laws of robotics was entirely irrelevant to the points he was trying to make.
I'd trust it more if it didn't have a fallible human behind the wheel.
I trust computers not to drink, drive sleepy, fiddle with the radio, talk to the hounddog on the CB, text cousin Willy, be aware of what was happening around it for a full 360 degrees every microsecond.
As opposed to the fallible human that programmed those computers... who have no personal responsibility over the vehicle, who may have been drunk, sleepy, fiddling with the radio, etc... Ya, I know - but testing! blah, blah, blah ... That still makes my ass twitch.
It must have been something you assimilated. . . .
I gather you don't have any knowledge of modern electronics including A to D conversion. Microseconds are an eternity.
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
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 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 Law.
Agreed that a Robot is no more a colleague than a screwdriver.
I think you're wrong about Asimov, though. It's obvious that to write about theoretical concerns of future technology, the author must proceed without knowing how to actually implement the technology, but may be able to say that it's theoretically possible. There is no shortage of good, predictive science fiction written when we had no idea how to achieve the technology portrayed. For example, Clarke's orbital satellites were steam-powered. Steam is indeed an efficient way to harness solar power if you have a good way to radiate the waste heat, but we ended up using photovoltaic. But Clarke was on solid ground regarding the theoretical possibility of such things.
Bruce Perens.
Neither humans nor computerized vision systems update every microsecond. 10 milliseconds is a representative number for a fast system.
Contribute to civilization: ari.aynrand.org/donate
what about repair and the push to keep the line moving as much as it can while working to repair the automation?
abolish health and safety laws?? and when the rich boss says do some unsafe thing or your gone some may just say fuck it and beat the shit out that boss just so they can get free room, board and doctor in the prison!
You haven't actually read Asimov's Robot stories, have you? If you had, you'd know that the implementation of the three laws of robotics was entirely irrelevant to the points he was trying to make.
I had read all of them - some several times - by 1965. When were you born?
I am sure that there are many other solipsists out there.
Oh dear. Like the other AC (or are you the same?) you don't seem to grasp that even if Asimov's laws failed to achieve their intended consequences, that has nothing to do with the fact that they could not be implemented anyway.
In other words, Asimov may have been interested in showing that the laws didn't accomplish what was intended. But they were also impossible.
One of the rules of hard science fiction is that the author must not introduce more departures from known scientific facts than absolutely necessary. Ideally, one departure should be the limit.
Asimov's "Positronic Brains" were a flat-out "deus ex machina". He assumed that the problems of both weak and strong AI would be solved; 50 years later neither has even been approached.
I am sure that there are many other solipsists out there.
But Asimov blandly ignored the obvious fact that there is no known way to implement such laws, which incorporate high-level abstract notions and moral principles.
He didn't ignore that, he simply assumed that it would become possible in the future -- and extensively explored the ways in which it could still go wrong, showing that even with that nonexistent technology, the seemingly foolproof Laws of Robotics were anything but foolproof.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
There is a rich and gory history of Lock Out Procedures that were followed and failed.
THEY ARE NOT FOOL PROOF.
After being almost killed by a Lock Out Procedure fault, I take an extra step.
Once electrical power has been Locked Out and Tagged, I install a substantial shorting bar across the incoming power lines.
If something ( or someone ) fails, there could be major damage to the plant electrical generating system - but my ass is safe.
I have never seen any employer recommend this step, probably because it is less costly to replace a worker than a power generation subsystem.
Can you trust humans not to be evil?
But Clarke was on solid ground regarding the theoretical possibility of such things.
I strongly disagree. There is absolutely no sign as yet that it will ever be possible to create robot brains that understand such high abstractions as "harm", or indeed even "human being".
I am sure that there are many other solipsists out there.
You're approaching it from an anthropomorphic perspective. It's not necessary for a robot to "understand" abstractions any more than they are required to understand mathematics in order to add two numbers. They just apply rules as programmed.
Today, computers can classify people in moving video and apply rules to their actions such as not to approach them. Tomorrow, those rules will be more complex. That is all.
Bruce Perens.
Microseconds are not important at this scale. There is nothing that physically can react that fast. Milliseconds is sufficient and for that there are hundreds of sensors that meet the requirements.
Stop anthropomorphizing machines. They hate that.
Have gnu, will travel.
I know a couple of truck drivers. From what I've been told, their job entails a lot more than just driving. Yesterday one was telling me problems they encountered when their truck was mis-loaded which I would assume is also a common problem.
They'll hire ACLU lawyers, be set free, join a gang, and go on fuel stamps.
Contribute to civilization: ari.aynrand.org/donate
Many years ago, our state highway department tested a drawbridge. The test involved switching the controls into a 'maintenance mode' and exercising the control logic without actually opening the bridge. Unfortunately, the maintenance mode selection failed and the bridge actually opened. During rush hour*, killing one woman when a bridge section suddenly popped up in front of her car. Investigations were done and it was found that the original control design did not include a comprehensive failure modes and effects analysis that would have revealed the passive fault.
*I don't know why they didn't test it at 2AM. A full end-to-end test with the barriers coming down and the bridge opening. They would have stopped about four cars for 10 minutes. On second thought, I do know why: Union rules. First shift operation costs less, even though the bridge operators are still sitting there for 8 hours.
Have gnu, will travel.
"The machine loaded the hardware onto Holbrook's head."
I'm not sure what the literary device is being used here, but that word choice really makes it the entire blurb leading up to this point makes it sound like the robot became sentient and was on a mission. The last line (the bit I quoted) was equal parts gruesome, horrific, and fantastic. I would read a book by whomever wrote the bit there.
But Asimov blandly ignored the obvious fact that there is no known way to implement such laws, which incorporate high-level abstract notions and moral principles.
Emphasis mine. What's the point of fiction set in the future if we can predict how everything would work out?
Ezekiel 23:20
No they don't.
source: Slashdot
A single car also has the potential to kill hundreds of people (cause a massive wreck on a busy freeway).
Sure it was. Unfortunately it could only understand a few dozen primitive opcodes.
A robotic arm has great mass and can move very quickly. Working around robots involves obeying safety rules and following procedures without exception. If a worker fails to follow the proper practices the effect of a robotic arm hitting a human head is about the same as a heavy golf club swatting the head at great speed. And workers do tend to disconnect safety systems to make working on the robots a bit easier. There are some trades in which one mistake is the last mistake one will ever make.
...might conclude that the death was anything but accidental.
Such is the fate of all who fail to bow down before our new mechanical overlords.
I've calculated my velocity with such exquisite precision that I have no idea where I am.
To hell with not being evil, unless it's the miracle software that has become truly self aware, that's not the problem.
The real problems are
1) Can you trust it to react sensibly to completely unexpected situations, mislabeled roads, etc.? (that's what's been behind the majority of self-driving car deaths so far)
2) Can you trust it's security to prevent it from being hacked and intentionally used as a weapon?
It's really seeming like 1 is a long way away, though perhaps we're about at the point where the elimination of human error at the point of control prevents more damage than the lack of sufficient AI competence causes.
As for 2 - given the general state of computer security, especially within the automotive industry, but basically everywhere, it seems obvious that the answer should be "hell no"
--- Most topics have many sides worth arguing, allow me to take one opposite you.
No one's claiming the robot is responsible for her death; they're claiming the manufacturers are responsible for her death, due to the robot's malfunction. The hyperbolic choice of language, I feel, purposefully crafts an idea in the audience's mind that's actually rather detrimental to the design and construction of robots that *can* analyze context and make autonomous decisions.
Not at all. Strong AI is not necessary to the argument. It is perfectly possible for an unconscious machine not considered "strong AI" to act upon Asimov's Laws. They're just rules for a program to act upon.
In addition, it is not necessary for Artificial General Intelligence to be conscious.
We have a lot to learn of consciousness yet. But what we have learned so far seems to indicate that consciousness is a story that the brain tells itself, and is not particularly related to how the brain actually works. Descartes self-referential attempt aside, it would be difficult for any of us to actually prove that we are conscious.
Bruce Perens.
The AC was being an ass, but he is making a valid point. Asimov wrote "I, Robot" 10 years before the existence of "computer programming" as we know it. The microprocessor didn't exist yet. "Programming" was more like electrical engineering than software engineering is today.
there is no known way to implement such law
Duh, of course! But to criticize Asimov as having "blandly ignored" the limitations of computer programming is insulting to his intelligence and detracts from the point of the books. They aren't engineering books about implementation. They are imaginations about the moral implications of what could possibly be. If anything, his books explore the moral ambiguity of the 3 laws of robotics, and talk about how hard it would be to write such laws given the assumption of a perfectly logical being that could interpret them. That's what makes it science fiction.
Um, I'll have to assume that you weren't around for April, 1968, when the leading AI in popular culture for a long, long, time was introduced in a Kubrick and Clarke screenplay and what probably should have been attributed as a Clarke and Kubrick novel. And a key element of that screenplay was a priority conflict in the AI.
Bruce Perens.
Just saying, what a perfect way to stage someones death. How could anyone know?
âoeTolerance applies only to persons, but never to truth. Intolerance applies only to truth, but never to persons.
... as an Uber Driver.
of coarse that is most likely but like the aritcle I though murder was a more titillating possibility. It is still a whole lot more likely then a machine 'mysteriously' moving and 'killing' a 'co-worker'. Honestly this is about as mysterious as someone being hit by a truck because they were working underneath it and the jack fell, but murder is at least a possibility and that what it would be on every TV show I've ever seen ;)
âoeTolerance applies only to persons, but never to truth. Intolerance applies only to truth, but never to persons.
I take it you've never worked on safety critical software which had to meet a certain standard of development and testing, e.g. mil 882 or iso 26262? For that stuff you'd have to have willfully malicious management for bad code to slip through in such a way that it could cause issues (and then they'd sure as shit be liable for it since the whole point of those standards is to leave a paper trail).
You don't seem to know the difference between input and output, but that being said, a Beowulf Cluster of Computers should do it. Someone should invent GPUs and massively parallel systems!
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
The word is derived from the Slavic language root meaning "work" or "worker", and strongly suggests that a robot is to some extent intechangeable with human workers.
Did you not see "The World's End"? A running joke in the movie is that antagonists are not robots as robot means slave ("... and believe me, they are not slaves").
The word is derived from the Slavic word "robota", which could mean "forced worker" (peasants "obligated to compulsory service" under feudalism, see Robot Patent) or slave.
In a civil liability suit, you sue everyone involved, from the part manufacturer to the installer to the operator to the owner of the space being leased to the guy that printed the sign out front...
Why? Because, if you don't, the easiest oath out of liability is for all of the parties of the suit to point to the unnamed party and snow a jury. You sue everyone so they do their due diligence to demonstrate a lack of fault in deposition and then get you pointed to the party at fault. This is not new or unusual. It's sadly necessary.
We have natural selection again, now just to make the robots more unsafe so we can avoid Idiocracy altogether.
there is no known way to implement such law
Duh, of course!
Why is that an, "of course"? The article describes a woman crushed in an assembly line. If a self driving car can recognize a human and apply the brakes, I think a similar algorithm could have put the assembly line robots into a safe mode.
Asimov blandly ignored the obvious fact that there is no known way to implement such laws, which incorporate high-level abstract notions and moral principles.
Which was the whole point.
His robot novels were parables about how even lofty goals like the Three Laws still left many loopholes leading to disaster.
A screwdriver is as much a colleague as Jack or Johnny (Daniels and Walker)
Don't fight for your country, if your country does not fight for you.
I agree that an algorithm could recognize a human and put the robots into a safe mode. But Asimov's 3 laws were much more complex than that.
Asimov's laws are more like an attempt to codify moral behavior. Humans can't even implement those laws. The laws required the robots to know the future and anticipate indirect consequences of their actions. In one of the stories, the robots implemented the laws by creating a religion. In another, the robots had to anticipate an object that might fall and indirectly crush a human, but then they also recognized that there was a hidden failsafe that would prevent the human from getting hit. The robots had to balance different conflicting goals and make a judgement between doing its job and saving a human. The movie, although not Asimov's writing, gave other examples like that. It showed a robot deciding which of two people to save, or a robot deciding between saving lives and preserving a human's freedom and right to decide.
The fact that human brains can do it strongly suggests that it's possible to make something artificial that can do the same thing. That doesn't necessarily mean we can do it with our current software approaches, or even in anything involving silicon, but it clearly is possible.
(With the pace of recent advances in AI I would argue that it seems likely to be doable in software, and probably very much sooner that you might expect.)