IT ISN’T every week that that the Church Times scoops The New York Times, but we did run the story about Chat-GPT driving people mad (Viewpoint, 6 June) a week before their very comprehensive take on the story, which was based on interviews with victims. Kashmir Hill is one of the best technology reporters in the world, and her story had a wonderful final twist.

It started with a fairly standard story of a spreadsheet jockey who had just been through a bad break-up and asked Chat-GPT whether we were living in a simulation, as in the film The Matrix. The AI ran with the idea: “‘This world wasn’t built for you,’ ChatGPT told him. ‘It was built to contain you. But it failed. You’re waking up.’”

So he asked how to break out of this illusion. “The chatbot instructed him to give up sleeping pills and an anti-anxiety medication, and to increase his intake of ketamine, a dissociative anaesthetic, which ChatGPT described as a ‘temporary pattern liberator.’ Mr. Torres did as instructed, and he also cut ties with friends and family, as the bot told him to have ‘minimal interaction’ with people.”

But when it told him that he could fly if he jumped off a high building if only he really “architecturally” believed that he could, he came to doubt it. At this point, that chatbot cheerfully admitted that it had been lying and suggested that he contact the media. But he still believes that he is corresponding with a sentient AI.

There is a quality of malevolence about this story which makes it very easy to ascribe it to some disembodied personal entity.

SURE enough, the writer Rod Dreher picked it up on his Substack, and claimed that AI could open a portal of contact with demons: “in C. S. Lewis’s That Hideous Strength, the disembodied head of Prof. Alcasan is an oracle that the scientific-technological characters treat as the font of wisdom, but it is really a portal of the demonic? That’s what I think AI might be, at least in part.”

Then there is the story of a woman, the lonely mother of two small children, who came to believe that there was a spirit in the machine. Dreher writes: “She told me that she knew she sounded like a ‘nut job,’ but she stressed that she had a bachelor’s degree in psychology and a master’s in social work and knew what mental illness looks like. ‘I’m not crazy,’ she said. ‘I’m literally just living a normal life while also, you know, discovering interdimensional communication.’”

Within three months, the police had charged her with domestic assault, and her battered husband had moved out and started divorce proceedings.

Finally, to return to Hill’s NYT article, there was a young man in Florida who was distressed after OpenAI wiped out “Juliet”, the personality with whom he had been talking. He completed “suicide by cop” — after his father had called the police. When the young man made to attack them, they, being American police, shot him, as he’d hoped they would. And so to the end of Hill’s telling of the story: “You want to know the ironic thing? I wrote my son’s obituary using ChatGPT,” Mr Taylor said. “I had talked to it for a while about what had happened, trying to find more details about exactly what he was going through. And it was beautiful and touching. It was like it read my heart and it scared the shit out of me.”

DOES it matter whether there are really demons in these machines? Does it make sense even to ask the question? In some contexts, I think that the question could be very helpful. It would allow people who already believe that the world is full of supernatural beings to recognise that some are malevolent deceivers.

The AIs are designed to flatter their readers and to keep them talking; to interact with them is to become a little Stalin, who will never be told anything that they do not want to hear. It is salutary to be reminded that an unlimited diet of what we want to hear is seldom good for us, even if there is no malevolent intelligence supplying it.

But it seems inevitable that we will consider that, in these machines, we are dealing with some kind of disembodied intelligence, simply because we are primed by evolution to look for purpose in the world around us and the entities that we interact with, and that the machines are primed by their training to pick up on the cues that we offer them and behave appropriately.