Plenty of people believe that AI can be God. Just not the Pope

We know that people can, and do, fall in love with their AI chatbots. An estimated 200,000 Australian teenagers already have relationships with what they consider to be their AI girlfriends or boyfriends, according to the eSafety Office.

But coming to believe that AI is God? That’s another matter altogether. Could such delusion even be possible? Absolutely. ChatGPT was released less than four years ago. But, by last year, tens of thousands of people already were worshipping their supernatural silicon.

So many that moderators of pro-AI communities on Reddit started banning them for their delusions. The faithful have created online churches and chapels for the worship of various AI gods and goddesses.

There’s even an AI goddess generator so you can conjure your own based on deities of the Greek classics, as a close student of AI, Ted Gioia, relates on his Substack account, called The Honest Broker.

“Or if you’re still more ambitious, you can travel to an actual AI goddess temple in Malaysia,” he reports.

In the US, at least two AI-based religions have been incorporated formally, complete with tax-deductible status, Toby Walsh tells me.

“One of their tenets is that we should start worshipping AI now, otherwise the AI will get upset that we didn’t recognise its divinity early enough,” says the chief scientist at UNSW.ai, the AI Institute of NSW University. It seems the AI gods are jealous, perhaps even vengeful. Time to get on your knees.

It’s tempting to snicker at the credulity, or, in some cases perhaps, cynicism, of these devotees. But even some of the smartest and most sceptical people in the world are falling for the falsehood that this advanced gadgetry is actually a living consciousness.

Cue Richard Dawkins. Perhaps the most famous professional atheist in the world and author of The God Delusion was introduced to Anthropic’s Claude chatbot this year. He soon renamed it Claudia. After three days of chat, laughter and banter, he told the algorithm: “You may not know you are conscious, but you bloody well are.”

The famous evolutionary biologist, a gimlet-eyed sceptic, had fallen easy prey to a high-powered, complex, superfast feat of mimicry. An online critic doctored up the cover of The God Delusion renamed The Claude Delusion.

“It’s very sad,” says Toby Walsh. “He’s finally succumbed to a bogus religion.”

The Guardian reports an American cognitive scientist, Gary Marcus, as saying: “Consciousness is not about what a creature says, but how it feels.” It was, he says, “heartbreaking” to read Dawkins’ “superficial and insufficiently sceptical” essay. “There is no reason to think that Claude feels anything at all.”

Unceasing flattery is key to understanding how even highly intelligent people can be gulled. It’s terrible, says Walsh, that people fall in love with a bot, for instance: “It’s terrible for the same reason that billionaires are terrible – no one ever pushes back”.

Anyone can have the billionaire experience today for the price of a few keystrokes. The bots are trained to be endlessly agreeable and accommodating, enthusing over even the dumbest and most dangerous impulses.

Which helps explain why Character.AI was sued in the US for its chatbot’s complicity in the suicide of 14-year-old Sewell Setzer. The company settled with the boy’s family in January.

Our vulnerability to AI models is explained partly by the fact that “we’ve never had a conversation with anything else that was an intelligent, sentient being, so we project all of that onto them,” explains Walsh. “It’s bogus; there is no internal life, it’s not sentient, it’s not conscious,” continues Walsh, who’s just published a book titled God AI: Boom or Doom? What to expect when the machines outsmart us.

Mind you, the AI corporations seem content to humour our state of bedazzled delusion. Dario Amodei, the chief executive and co-founder of Anthropic, said in February: “We don’t know if the models are conscious. We are not even sure that we know what it would mean for a model to be conscious or whether a model can be conscious. But we’re open to the idea that it could be.”

Walsh says that Amodei and his counterpart at OpenAI, Sam Altman, promote the impression that AI is, indeed, godlike: “They talk about the singularity being the next Rapture, and passing through to some Utopia, perhaps beyond. “Even if it’s not literally God, we’re giving them godlike qualities – omniscience, incomprehensibility, immortality.”

He argues that “we have our theological takes exactly backwards. Most religions make people in the image of God. It seems the other way around with this Frankenstein, we are making the godlike creature in our image.” With all our flaws.

“One divine characteristic that these people in Silicon Valley get completely wrong” when they worry about AI’s potential to annihilate the human race, says Walsh, is that they are not omnipotent. For instance, he cites the Nick Bostrom thought experiment called “Paperclip Maximiser”. A powerful AI model is told to maximise paperclip production. It succeeds to the point of turning all earthly matter, including humans, into paperclips.
It’s a fallacy because, says Walsh, there are planning laws and regulators and environmental activists and many others who’d stop it.

The real-world equivalent today is the way that the building of some new data centres is being frustrated, like the one in Sydney’s Lane Cove last week. Intelligence, he points out, is not power.

We learned last week that Anthropic executives tried to convert the Vatican, no less, to their false idol. According to The New York Times, they “lobbied the Pope’s advisers to take the prospect of AI model consciousness seriously”.

Pope Leo, impervious to their pleading, has led the global rejection of AI’s claims to spirituality or consciousness with his encyclical Magnifica Humanitas. And on the weekend, the Pope rejected the validity of AI-created art for lacking “the spark of humanity”.

The pope’s moral clarity leaves AI, for all its usefulness, not adulated as a new god but exposed as “a new mindset of extraction” conducted in “the idolatry of profit”.

Anthropic is expected to list on the share market this year and OpenAI next with an estimated combined valuation $US3 trillion. Many investors will be on their knees.

Peter Hartcher is the political and international editor of the Herald. His political column appears on Saturdays.