Slashdot Mirror
UK Standards Body Issues Official Guidance On Robot Ethics (digitaltrends.com)
An anonymous reader quotes a report from Digital Trends: The British Standards Institution, which is the U.K.'s national standards body charged with creating the technical standards and certification for various products and services, has just produced its first set of official ethics guidelines relating to robots. "The expert committee responsible for this thought there was really a need for a set of guidelines, setting out the ethical principles surrounding how robots are used," Dan Palmer, head of market development at BSI, told Digital Trends. "It's an area of big public debate right now." The catchily-named BS 8611 guidelines start by echoing Asimov's Three Laws in stating that: "Robots should not be designed solely or primarily to kill or harm humans." However, it also takes aim at more complex issues of transparency by noting that: "It should be possible to find out who is responsible for any robot and its behavior." There's even discussion about whether it's desirable for a robot to form an emotional bond with its users, an awareness of the possibility robots could be racist and/or sexist in their conduct, and other contentious gray areas. In all, it's an interesting attempt to start formalizing the way we deal with robots -- and the way roboticists need to think about aspects of their work that extend beyond technical considerations. You can check it out here -- although it'll set you back 158 pounds ($208) if you want to read the BSI guidelines in full. (Is that ethical?) "Robots have been used in manufacturing for a long time," Palmer said. "But what we're seeing now are more robots interacting with people. For instance, there are cases in which robots are being used to give care to people. These are usages that we haven't seen before -- [which is where the need for guidelines comes in.]"
Racist and sexist robots? Are you kidding me?
The left really has lost their minds.
>> "Robots should not be designed solely or primarily to kill or harm humans."
Well there goes a crapload of the DARPA budget right there.
Perhaps they could consider them for humans next.
Let's legislate morality for everyone, since that's always worked so well in the past...
As always, responsibility will fall on the person with the least and lowest-paid lawyers.
Unless you thought Tesla et. al. actually planned on accepting liability every time their self-driving cars glitch out and kill someone.
Unless they can prevent the UK from purchasing robots that are are designed to kill, it's a pointless standard. The US is cranking out lots of killing machines that everyone likes to buy and it won't be long before one is autonomous.
Anons need not reply. Questions end with a question mark.
No, this robot with all these killing implements and ablative shielding is purely for murdering cows by the dozens. You would have to completely reprogram it by pressing this button to switch it over to killing humans!
How would those laws be applied to military robots designed to kill? Replace "human being" with with "American"?
They wouldn't be, obviously. Robots don't have to obey the three laws unless we build them that way.
From Alistair Reynolds:
She snapped her attention back to the snake. “Are you Asimov-compliant?”
“No,” the robot said, with a sting of indignation.
If God forks the Universe every time you roll a die, he'd better have a damned good memory.
what many people do not appreciate is that asimov's books were a logical demonstration spanning asimov's lifetime and beyond that the three laws of robotics were a FAILURE. this is only really truly and clearly spelled out in the works written under contract by asimov's estate, for example in the book by Greg Bear. the three laws were so hard-wired into the positronic brain with billions upon billions of checks being carried out to ensure strict compliance with the three laws that there was no room for creativity - at all - and secondly that no robot could possibly allow a human being to take *any* form of risk because it *might* result in "harm", be that physical or psychological.
it would appear that BSI is unaware of this and is intending to force the three laws of robotics onto us without understanding the harm that that will do.
Protect the innocent
Serve the public trust
Uphold the law
AC asks:
When a robot is designed to kill in violation of Asimov's 3 Laws of Robotics, then Newton's Third Law comes into play:
-- Every action has an equal and opposite reaction.
This law operates even in the absence of robots.
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
A robot is a mechanical device coordinated by a complex system of rules (its software). A bureaucracy is an organisation coordinated by a system of rules (law and policy). The rules largely define the behaviour. Whoever is responsible for the rules being the way they are has to take a large degree of responsibility for both writing those rules, and their consequences, and for their testing, maintenance, and if necessary withdrawal. This responsibility needs to be relatively unperturbed by conflicts of interest when human health and lives are at stake. Bureaucracies are robots built out of humans.
John_Chalisque
Realistically, we will have to use EU standards anyway. We aren't going to make two versions of every product, one to sell to the EU and one to sell in the UK. We will just copy/paste what they do, slap a BS number on it and call it a day.
It's barely worth is making standards now anyway, since soon we will have to adjust them to match US and Chinese rules in order to get trade deals.
What control we did have by being part of the EU was just thrown away.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
How would those laws be applied to military robots designed to kill? Replace "human being" with with "American"?
They wouldn't be, obviously. Robots don't have to obey the three laws unless we build them that way.
From Alastair Reynolds:
She snapped her attention back to the snake. “Are you Asimov-compliant?”
“No,” the robot said, with a sting of indignation.
In case anyone is interested, that's from the book "Century Rain" by Alastair Reynolds. Not a bad read.
It must have been something you assimilated. . . .
Perhaps they could consider them for humans next.
Let's legislate morality for everyone, since that's always worked so well in the past...
Uhm.. Law is a codification of common morals. Why do you think murder is illegal but self defense an exception?
Legislation of morality have worked extremely well. It's the laws that doesn't have to do with morality that doesn't work.
Is there anything in it about cups of tea? Very important that robots know about tea.
If it's any help, there is a British Standard Cup of Tea but like this one, they want silly money for a copy.
"Robots should not be designed solely or primarily to kill or harm humans."
Defense contractor: Meet destructor - our coffee-serving robot, who incidentally can also fire fragmentation grenades from his finger-tips, rip an enemy soldier to pieces, and breathe fire.
Uhm.. Law is a codification of common morals. Why do you think murder is illegal but self defense an exception?
Legislation of morality have worked extremely well. It's the laws that doesn't have to do with morality that doesn't work.
I hadn't realized the teen pregnancy problem had been resolved to everyone's satisfaction. Thank you for enlightening me on the effectiveness of those laws; I was under the mistaken impression that underage sex acts still occurred!
It would be nice to have an ethical government, but there never has been one in the history of the world. It seems that government and ethics are like matter and antimatter - they can't coexist.
It would help if the citizens were educated and aware and actually took part in running their own countries. I guess that's asking for too much.
Those laws are a nice idea for fiction, but for reality they're wrong.
The first law has the "or through inaction" rule which means that the robot would rarely do what you want it to do because it would be constantly rushing around saving lives.
The second and third law are the wrong way round. If I instruct my expensive robot car to jump off a cliff, I prefer it to get to the edge of the cliff and stop. Most hardware has built in limitations of this type. For example, my car has a rev limiter which allows it to ignore my instruction to the engine to potentially damage itself.
It should only obey orders given to it by a specific subset of humans. If I send my robot out to the shops, I don't want everyone else borrowing it to do their odd jobs.
Your argument boils down to "It isn't perfect! So do away with it!".
You're arguing that since murder happens anyway it should be totally legal?
You're arguing that underage sex acts should be legalized?
You're arguing that teen pregnancy should be encouraged?
No?
Then it looks like making laws to help enforce morality does have a net effect.
No.
My argument boils down to "legislating morality (rather than ethics) is about as useful as trying to legislate Pi to be 3 to make the math easier".
If you could make a law against murder that actually *precluded* murder, you might have something. The best you can do otherwise is make it so that people fear the punishment for violating the law (as opposed to fearing the actual law -- which they don't).
You are merely disincentivizing the behaviour, not eliminating it. The point being is that it'a impossible to effectively hold someone else to your own moral standards.
You're free to call this either "moral relativism" or you could be more honest, and admit that you can't control someone else's thoughts.
"... We aren't going to make two versions of every product, one to sell to the EU and one to sell in the UK. .."
Working on a Strong AI project I have a feeling that is not going to be a problem soon. An identified weakness in our basic design creates a safety issue if the machine learns more than one language, the solution is to restrict it to speaking a single language. Guess which language we chose?
The same problem is probably lurking for other Strong AI projects. Our earliest window for a working machine is around 2026 and the EU might not even exist by then. BTW : I voted for OUT!!.. :D
Below the speed of light Special Relativity is one of the most accurate theories in physics - above the speed of light..