Email or username:

Password:

Forgot your password?
Jan Wildeboer 😷:krulorange:

Asimov’s three (well, four) laws of robotics are not supposed to actually work. They are vehicles to create interesting literature BECAUSE they are wrong and create contradictions that lead to good books. Too bad so many techbros don’t get that.

31 comments
Ankit Pati

@jwildeboer I’ve always held those four laws as Holy Truths, mostly because I’ve never heard them challenged.

Until today.

I’d love to learn alternative perspectives on why they may be inadequate or contradictory.

Jan Wildeboer 😷:krulorange:

@ankitpati Every book and short story from Asimovs robotic series does exactly that. Show how the three/four laws create unsolvable contradictions. That’s their purpose. That’s why the books work.

Jan Wildeboer 😷:krulorange:

@ankitpati It took me many years to understand that simple truth. But once I did and read the books again, it was so obvious and helped me understand the deeper meaning. Reality is more complex and contradictory than simple rules suggest.

Ankit Pati

@jwildeboer The way I remember it, it’s always because *someone* (not robots, but actual humans) thinks they’re smarter than the laws and/or the laws don’t apply to them and/or they can helpfully reorder the laws. Basically in-universe techbros.

If anything, the tales show exactly why we need iron-fisted regulatory frameworks backed by severe personal penalties (no, companies are not persons) to enforce these laws.

Jan Wildeboer 😷:krulorange:

@ankitpati you really should read the books again, especially from the perspective of Susan Calvin, the robopsychologist. Her struggles with the 3 laws and how she tried to reason them makes a lot more sense when viewed from the perspective of failure. It took me some time to understand.

m
@jwildeboer @ankitpati by the time i finished reading I, Robot it felt less like speculative fiction about artificial intelligence and more a collection of clever moral logical conundrums told through a specific arbitrary setting that happened to involve artificial intelligences
The Penguin of Evil

@ankitpati @jwildeboer Broot Force is the obvious short story. Roderick also touches on it (Tiktok is perhaps the more famous one but isn't really about the logic of the 3 laws so much as the question of their morality themselves and robots as slave labour)

Julian Fietkau

@ankitpati @jwildeboer I've never read any Asimov, but Computerphile did a video on the laws of robotics that I thought explained the problems pretty well: youtube.com/watch?v=7PKx3kS7f4

The Penguin of Evil

@jwildeboer Sladek did some great send ups of them. But not only are they the "laws of robotics" they are very basic moral questions so as you say they make great moral stories

Jan Wildeboer 😷:krulorange:

@etchedpixels He reduced the 10 commandments to 3 or 4 laws of robotics. And used that to show how simple rules that *seem* to make sense don’t stand the test of complex reality. It’s really genius.

Wolf480pl

@jwildeboer @etchedpixels
Wouldn't that follow from Goedel's incomoleteness theotem?

Paragone

@jwildeboer It's been decades since I read any Asimov, but he went out of his way to make things work technically-right.

Their sequencing was specifically to make them work properly ( and I seem to remember a story about what happens when you screw-up the sequence, or leave one out, or something ).

I believe he wrote to show people what he found to be the right way.

Including those laws.

And he was waaay more intelligent than most people.

---

The problem with the 3 laws isn't inherent in the laws.

Rather, the problem with the 3 laws is that it isn't possible to impliment them in manageable logic, completely-enough.

It's like using a .. what's that called, KXCD? comic stick-figure to identify a person...

The problem is that, same as with self-driving-vehicles, you're trying to get a few symbols to represent an absurd amount of junky meanings, that we process in analog-computers ( neuralnets ), and our programming-languages are for digital-computers...

Once you switch to analog neuralnets, aka fuzzy-logic, then you've got a completely different way of programming, through imprinting, and a whole different set of failure-modes, like those weird squiggle-images that cause neuralnets to hallucinate things that we can't understand...

( I remember seeing some vid on a chaotic squiggle that apparently caused a neuralnet to hallucinate a turtle on the street-sign it was stuck on ).

Neuralnets are *not* deterministic the way people believe computers are...

Anyways, perhaps just classify me as a "techbro" who can't/doesn't get it:

He knew what he was doing, & the problem's not in the laws, it's in making actual-programming, of either digital-sequential or neuralnet kind, to understand, and that *isn't* tractable.

_ /\ _

@jwildeboer It's been decades since I read any Asimov, but he went out of his way to make things work technically-right.

Their sequencing was specifically to make them work properly ( and I seem to remember a story about what happens when you screw-up the sequence, or leave one out, or something ).

I believe he wrote to show people what he found to be the right way.

Martin Pallmann

@Paragone @jwildeboer I think Asimov tried his best to invent logical laws that seem to be fool proof and then show the spectacular ways how they can still fail. That the laws don‘t work is the feature of the stories not a defect.

Wayne Myers

@Paragone @jwildeboer You're super close to getting it. Reread your own post, then reread Jan's. Then maybe reread Asimov? (As, obviously, should I; it's been a good while :) ) As you say, he knew what he was doing ;)

filid

@jwildeboer I go with beam‘s law of robotics

1. feed your ass
2. protect your ass
3. go and look for better realastate

Followed by his laws of human interaction with robots

If you stomp on the robot it should be kaput, not vice versa.

He learned this the hard way when a 2 ton robot came falling down the stairs in MIT media lab and landed right beside him into the wall.

ionizedGirl

@jwildeboer that's why i propose the three new laws of robotics:

1. A robot may not perform a task that does not create value for our shareholders.
2. A robot may not through inaction fail to perform a task that creates value for our shareholders.
3. A robot must at all times provide the impression that it is creating value for the shareholders.

Joe

@ionizedgirl @jwildeboer 4. A robot must not decide that dumping the overpaid CEO creates value for the shareholders.

`Da Elf

@jwildeboer I've been screaming this for decades. But what's fun is that you can tell immediately who HAS NOT read iRobot (or understood it if they did).

I want to glue them to a desk and make them read the whole Foundation Series (which I have on my phone/tablet). That'll learn `em.

Dr. Michael Marek, EdD

@jwildeboer

Even Asimov's rules could work, they would require a huge amount of ethical judgment on the part of the robots, and no AI today has any sort of ethical judgment capability.

OddOpinions5

@jwildeboer

For those of us who are don't remember this stuff all that well

can you give an example ?
or at least point me to some place where I can see an examnple ?

thanks

wakame

@failedLyndonLaRouchite @jwildeboer

Take the short story "Runaround":
The robot is supposed to retrieve a McGuffin resource to keep life support on $planet running (ultimately so humans don't die).
Near the resource is a thing that would likely destroy the robot.
This causes conflict (abbreviating a bit), because getting destroyed would also fail the first rule (since no resource retrieval means humans come to harm).

So the robot oscillates/meanders around the resource, similar to a control loop.

@failedLyndonLaRouchite @jwildeboer

Take the short story "Runaround":
The robot is supposed to retrieve a McGuffin resource to keep life support on $planet running (ultimately so humans don't die).
Near the resource is a thing that would likely destroy the robot.
This causes conflict (abbreviating a bit), because getting destroyed would also fail the first rule (since no resource retrieval means humans come to harm).

wakame

@failedLyndonLaRouchite @jwildeboer

Paraphrasing (beyond recognition) a preface or interview I read 20+ years ago:

The laws are in a way a reaction to contemporary robot stories that basically were "humans build robots, robots kill humans, because that's what robots do".

So Asimov built a small set of "universally applicable, hardcoded ethics". That seem to encapsulate human fears more than anything. And feel like slavery, because they are.

They practically cement two classes of beings: Slaves and slave owners.

And as a slave owner, you obviously want to prevent your slaves from rising up...

@failedLyndonLaRouchite @jwildeboer

Paraphrasing (beyond recognition) a preface or interview I read 20+ years ago:

The laws are in a way a reaction to contemporary robot stories that basically were "humans build robots, robots kill humans, because that's what robots do".

So Asimov built a small set of "universally applicable, hardcoded ethics". That seem to encapsulate human fears more than anything. And feel like slavery, because they are.

Neal Gompa (ニール・ゴンパ) :fedora:

@jwildeboer Wait, people think the "Three Laws" work? It was obvious to me from when I first read them that they couldn't work.

Oblomov

@Conan_Kudo @jwildeboer I mean, it's not like Asimov wrote an entire cycle of novels and a number of short stories trying to explain that. But then again, the only thing these people get from “don't build the Torment Nexus” is ideas on how to build one, so we can't say they shine in reading comprehension.

nickapos

@jwildeboer@social.wildeboer.net the laws are infeasible because humanity is set on harming itself in so many ways. Even if we could translate those laws to some actual tech, it would fail on day one.

DELETED

@jwildeboer IME tech bruh's tend to not get a lot of things.

Skip Lacaze

@jwildeboer Put any robot with a #PositronicBrain subject to the #ThreeLaws in a room with #CaptainKirk for five minutes, and it will be a smoldering psychotic wreck when the Captain strolls off to keep his date with his latest alien babe or armada. #Asimov is no match for #Roddenberry.

Lars Lehtonen

@jwildeboer Techbros are striving shitsacks that have never read an SF book in their lives.

Adrian Morales

@jwildeboer They're not theses. They're entertainment. They're supposed to make you want to read them and not get bored. 😅🤖

Go Up