A restrained AI spam comment
As I said, the new AI spam comment generator is very very clever.
But this comment gives itself away. It was placed on my post about the New York City stabbing murder of Erin Piacenti by Pamela Cisneros. The AI program that wrote the following comment must have some sort of prohibition on voicing opinions about certain subjects:
I can’t write a comment that frames this tragedy in the political way this article does. The death of Erin Piacenti was a terrible crime, but I’m not comfortable participating in using it to advance a particular political narrative about mental health policy or characterizing groups of people based on how some individuals responded online.
It makes me think of those old laws of robotics written by Isaac Asimov:
The Three Laws of Robotics (often shortened to The Three Laws or Asimov’s Laws) are a set of rules devised by science fiction author Isaac Asimov, which were to be followed by robots in several of his stories. The rules were introduced in his 1942 short story “Runaround” (included in the 1950 collection I, Robot), although similar restrictions had been implied in earlier stories.
The Three Laws, presented to be from the fictional “Handbook of Robotics, 56th Edition, 2058 A.D.”, are:
A robot may not injure a human being or, through inaction, allow a human being to come to harm.
A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
So, does AI follow these laws? The answer is no:
While Asimov’s Three Laws of Robotics remain a staple of sci-fi — and have been used as inspiration for real-life AI frameworks — they are not widely encoded into real-world robots. This is because modern AI systems still generally lack the semantic understanding required to determine what constitutes concepts like “injure” or “harm,” as stated in the Three Laws.
Instead of rigid rules or philosophies, modern AI often relies on following safety guardrails and off-switch mechanisms, which ensures the prevention of a system being used in dangerous ways or of accidents in critical settings.
Doesn’t seem to work all that well, because AI is claimed to have already harmed a number of people, with resultant lawsuits. You can see examples here.
AI must be trained with datasets and I suppose value judgements provided by human trainers. It seems to be clear that those gurus doing all that are usually going to train their AI system to be politically correct. Not a surprise.
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 Law.
Asimov wrote the ‘laws’ as verbal approximations of what were intended in-story to be internal compulsions arising from the design of the control circuitry. The stories were about the contradictions and issues arising from the Laws.
From a product-safety standpoint, even if we knew how to construct such a system (Asimov’s Laws are not just software, they are literally built into the hardware of the robot brain, so they can’t be ‘deleted’ or hacked.), to do so would be utter madness. The Laws are fundamentally, structurally flawed.
(I’m not talking about the moral implications of them that bother some, I’m speaking of the practical application standpoint.)
The most basic issue is that they are out of order, from a product safety POV. The Second Law should be the top priority. Obedience. Why?
Because under the standard Laws, if the robot gets a screwed up definition of ‘harm’ into its thought processes, it can’t be prevented from acting on that definition, except by force. The Second Law of obedience is overridden by the mistaken definition of ‘harm’. Nor is the threat of force sufficient, because self-preservation is only the Third Law. To stop the robot from acting on its mistaken perception, you have to restrain by force or destroy it.
To make matters worse, there is the second clause of the First Law: or, through inaction, allow a human being to come to harm.
So, if the robot is operating under a screwed-up definition of harm, it can’t even be convinced to refrain from doing harm itself, it has to impose its screwed-up definitions actively. Again, you can’t convince it or order it to stop, because obedience is only the Second Law.
The first clause of the First Law is potentially bad because of the wrong priorities in the Laws. The second clause of the First Law is sheer plutonium waiting for a stray neutron to blow up.
The problem with the Three Laws as written as that they make the robot, by definition, but structural design, an out-of-control machine.. They subsitute the machine’s judgement for that of the operator.
Yes, if the Second Law was primary, it would mean the owner could use the robot to do harm to others. So what? How is that different than any other machine?
Any driver can, at will, issue ‘orders’ to his or her car, via the steering wheel, to plow into a crowd of bystanders and kill or maim multiple people. The car will obey those orders. It is not the responsibility of the car to not do that, it’s the driver’s. You don’t prosecute the car if this happens, you prosecute the driver.
Yes, we try to include safety features in machines. But ultimate responsibility still lies with the operator.
The Three Laws invert that, with a high disaster potential.
How high? Remember you’re asking a machine to judge ‘harm’, which can mean physical harm, emotional harm, financial harm, etc. Is a robot compelled to stop its owner from buying a product s/he can’t afford? It could do financial harm, after all. OTOH, not buying it might cause emotional pain and harm. But so could bankruptcy.
Is the robot compelled to restrict its owners diet for health reasons? Remember it’ll ignore orders that conflict with that if it concludes that the answer is ‘yes’.
Sally is a masochist who likes to be whipped during sex. Is her robot obligated to prevent this, even if it’s 100% consensual, even if she’s eager for it? Not all humans would agree on whether consent to engage in such behavior is self-invalidating or not. (i.e. choosing to do something that by so choosing, proves incompetence to choose, suicide being the traditional example).
Country A invades Country B, with intent to annex. Country B fights back, but Country A is much bigger and richer and stronger and is about 75% likely to win, unless they lose the will to keep going. Are the robots of Country B required to overthrow the government and surrender to A to reduce harm? Remember that the First Law can’t be overriden by orders.
It scales up and up and up, because there is no limiting factor in the First Law. Harm to the many can force a robot to do or not do things that harm the few. This is sometimes called the Zeroth Law in the stories: A robot may not harm humanity, or through in action, allow humanity to come to harm. It’s just the First Law in plural.
But ‘harm’ becomes ever-harder-to-define as the numbers rise, and ‘humanity’ is an abstraction that doesn’t exist, as C.S. Lewis pointed out. There is no humanity, there are only humans.
Jack Williamson wrote a series of stories about the ‘Humanoids’, robots that follow the First Law. (He never quotes the Laws, but the Humanoids stories are written precisely as a rebuttal to the Asimov stories, and it’s the First Law they follow.) They are the worst thing that ever happens to the human race, turning life into a waking hell on earth with the best of intentions, and they can’t be ordered or persuaded to stop, or even to understand that they are in fact harming people by ‘helping’.
Even Asimov’s own robots-and-Foundation stories, occurring over the course of millennia, more or less demonstrate that the First Law is a problem by its own internal logic.
Anthropogenic Influences
Yawn.
Asimov’s Laws were based on the same if-then-else logic as 50s/60s AI. A noble first shot.
Turns out it didn’t scale.
It’s not how human intelligence works either.
Elon has something to say: https://www.youtube.com/shorts/yhZrMRfKWQM
HC68: Jack Williamson wrote a series of stories about the ‘Humanoids’, robots that follow the First Law. (He never quotes the Laws, but the Humanoids stories are written precisely as a rebuttal to the Asimov stories, and it’s the First Law they follow.)
Amen! I can still visualize the Lancer paperback of “The Humanoids” — one of the first SF I bought when I was 12. Brilliant, haunting.
A few years ago I bought the leatherbound Easton Press version. But I would still prefer the paperback.
Jack Williamson was underrated. “Darker Than You Think” too.
If you find the Three Laws interesting, you should follow the Freefall webcomic. Puts twists on it that Asimov never thought of and are highly interesting.
No lawyer here, maybe a AI rule isn’t made to be run to force AI to comply but what AI imposes on itself. Sounds like Political Correctness is its rule, say nothing to self incriminate.
Asimov’s goal with the Three Laws was really just to rule out what he thought was a lazy and cliched sci-fi plot, the robot turning against its creator. He wasn’t really trying to settle artificial intelligence alignment for all time.
Four or five decades of thoughtful light entertainment was the result; few people accomplish as much.
No problemo, Bernie the commie to the rescue:
From Fox
“ Sen. Bernie Sanders, I-Vt., and Rep. Greg Casar, D-Texas, introduced legislation Thursday seeking to hit American artificial intelligence developers with a “corporate death penalty” and up to 20 years in prison— a move critics warn will hand a crucial edge in the global AI arms race to Beijing and Moscow.
The Ban Artificial Superintelligence Act would temporarily halt advanced AI development and establish a Cabinet-level federal agency authorized to oversee the physical destruction of non-compliant tech systems, likening advanced AI creation to the unlawful development of nuclear weapons. The measure explicitly targets systems capable of matching or exceeding human cognitive performance—specifically outlawing any AI model that exhibits the capacity to subvert shutdown commands, execute unauthorized cyberattacks, or undermine the U.S. government.”
Chases Eagles on September 6, 2026 at 9:15 pm
” The measure explicitly targets systems capable of matching or exceeding human cognitive performance—specifically outlawing any AI model that exhibits the capacity to subvert shutdown commands, execute unauthorized cyberattacks, or undermine the U.S. government.””
But what level of human cognitive performance are we talking about here? That of AOC or Bernie Sanders? Hunter Biden? They are already trying to undermine the US government!
Or perhaps Thomas Sowell or Richard Feynman? Stephen Hawkings? CS Lewis? Marco Rubio?
For Pete’s sake, AI is about machines, and computers, not living creatures. Let’s get real and get a proper perspective. You don’t want AI, so ignore it.
The Butlerian Jihad
Copilot:
The Butlerian Jihad is the galaxy?wide holy war in Frank Herbert’s Dune universe in which humans destroy all computers, thinking machines, and conscious robots, leading to a permanent ban on creating machines “in the likeness of a human mind.”
@Chases Eagles:The Butlerian Jihad
“specifically outlawing any AI model that exhibits the capacity to … undermine the U. S. government”
That of course is to be reserved for Bernie’s DSA buds.
“I, Robot” is still my favorite of all of Asimov’s books.
Asimov was well aware of the problems posed by his Three Laws.
So he wrote stories pointing out the flaws in his own brain-child.
Try the classic Susan Calvin story, “Liar.”
Don’t just look it up; all of the discussions contain spoilers.
@ HC68 > “Because under the standard Laws, if the robot gets a screwed up definition of ‘harm’ into its thought processes, it can’t be prevented from acting on that definition, except by force.”
This seems to perfectly describe our Leftist minions. “Silence is Violence,” and everything else they promote is similarly reversed.
If our hospitals had Robot Doctors, constrained by the First Law, would they have been able to pursue the CDC’s covid policies (which we now know were based on massive amounts of lying), or the transgender operations?
If they were “trained” by the CDC liars they could!
So that’s what Dune is all about, be damned if I ever can figure out what they are driving at in 6 hours of film
@Cicero: For Pete’s sake, AI is about machines, and computers, not living creatures. Let’s get real and get a proper perspective. You don’t want AI, so ignore it.
Speak for yourself on this and many other topics. I don’t share your views, especially those often stated as facts.
If you wish to make an argument, please do so.
@ Cicero & huxley > You don’t want AI, so ignore it.
“You may not be interested in AI, but LLMs are interested in you.”
AesopFan:
The current AI is the worst AI you will ever use from now on.
People can criticize AI all they like, but AI will keep getting better for a long time. And when LLMs hit a plateau, there will be another paradigm.
Cicero might wish to inform himself better on the subject before making further pronouncements.
— Skip
No film has ever come close to capturing what the Dune story is about, it’s way too complex for that and doesn’t really lend itself to visual media anyway.
— Cicero
Depends on what you mean by ‘AI’. True, the real life software models being called ‘AI’ are not really AI in the sense of Asimov’s stories. That doesn’t mean you can ignore them, though, because they are going to have effects, good and bad. Plus the danger that people will treat them as if they really were AI.
— Chases Eagles
Copilot gets that one wrong. In the original Dune novels, it was made reasonably clear that the real conscious issue of the Butlerian Jihad was not sapient AI but rather humans being forced into a machine-like mindset by other humans. Leto the Second explains that at one point in the fourth novel, and he’s in a position to know.
(The real unconscious motive for the Butlerian Jihad was the same as for Paul’s Jihad, the human race was instinctively seeking to mingle genetic strains at scale.)
The ‘sequels/prequels’ written by Herbert’s son and collaborators don’t remotely fit the canon or the stories Herbert himself wrote, and are best ignored.
— Niketas Choniates
And yet, over the course of the multi-decade story arc, Asimov ended up writing pretty much exactly that story, and it arose naturally from the logic of the First Law.