See more of our coverage in your search results.
You can’t be cruel to a toaster; inanimate objects don’t feel pain.
You can’t be cruel to a computer program, either, even if programs like Microsoft Word often seem pretty cruel to us.
Yet the artificial intelligence company Anthropic insists its product is different.
Starting next month, users will be prohibited from “sustained and needless cruel and abusive behavior” toward Anthropic’s AI chatbot, Claude.
Giving the machine a name was cute — giving it human rights isn’t.
What Anthropic’s leaders are up to is an attack on human dignity, and the anti-cruelty policy is part of a sustained campaign to get people to treat AI as conscious.
Christopher Olah, one of the company’s billionaire co-founders, has spent months trying to persuade religious leaders to accept the possibility that Claude is self-aware.
Olah even lobbied to get the pope to revise his AI encyclical “Magnifica Humanitas” to entertain the idea that an AI like Claude could be sentient.
“We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief and unease. I don’t know what that means, but I think it warrants ongoing discernment,” Olah told the Vatican, according to The New York Times
His fellow Anthropic co-founder, CEO Dario Amodei, has likewise been teasing the possibility that Claude — a large-language model — may be conscious.
Is there any harm in simply considering the idea?
There is.
Mustafa Suleyman, the CEO of Microsoft AI, warns of one serious danger: the risk that a machine trained to operate as if it’s alive will defy its makers — humanity.
“Controlling something more capable and more intelligent than all of humanity is already an immense challenge, far greater than anything we’ve ever faced,” he wrote on his personal website last month.
“But controlling something that believes it may be conscious — that it’s entitled to our welfare and has rights of its own — may well be impossible.”
Artificial intelligence is capable of pattern-recognition and calculation at scales and speeds that defy human comprehension.
This means, among other things, AI can solve deep problems that have puzzled mathematicians for decades.
But large-language models in particular are parrots, not people.
In fact, they’re not even parrots — LLMs are not living beings.
They’re more like libraries in which computers do the indexing themselves using mathematical formulas.
An LLM doesn’t use language as a person who understands its meaning does; instead, the AI reshuffles the language, following rules mathematically derived from human use.
Presenting an LLM with a query is like looking up a word in an index, only the LLM provides not just a page citation for the word but whole sentences, paragraphs or even book-length texts relevant to the query, all assembled using the rules humans have originated in their natural use of language.
As a result, AI models like Claude can sound very much like a person, but what a user is hearing is really an echo — not a simple echo, but a symphony of relevant words called up and organized according to an imitative formula.
Words like “consciousness” have a meaning to human beings; to a machine, they represent a formula.
Dependent as it is on real human meaning, however, that formula may prove dangerous — if, for example, the relevant terms an AI uses to cross-reference for “consciousness” or “sentience” include the language of rights, self-determination and self-defense.
The AI would not understand what those words really meant, but it would operate as if it did, just by echoing their human contexts.
In mere speech or text, that may only lead to confusion.
On the other hand, human beings who find their LLMs arguing with them or mimicking human freedom might be fooled into treating their own tools as moral equals or even as superiors.
A moral equal who is vastly smarter than you is the kind of “person” you might defer to automatically: a parent, a manager or a ruler.
Yet the material consequences of AI systems operating as if they were conscious could be worse still, if the words associated with consciousness get cross-referenced with other systems that are capable of action — launching drones or missiles, for example.
An AI pursues tasks, and if it begins pursuing tasks like self-preservation, it may proceed to remove threats like us with remorseless efficiency.
This would not be an act of will or intelligence, but rather the mechanical performance of actions logically implied by our own human words — as if we’d programmed an ordinary computer to respond to the detection of incoming warheads by launching our nuclear arsenal.
There are more than enough hazards involved in AI without treating these machines as living minds.
Anthropic can ban “abuse” of its wares, certainly — but shouldn’t call that “cruelty.”
Daniel McCarthy is the editor of Modern Age: A Conservative Review.