In the centre of Durham last week, at the far end of the old indoor market, my son and I hit pay dirt. Down past the fishmonger and the cleaning supplies, past the dead-eyed teens suffering through their summer jobs, right against the market’s back wall: Jim’s Vintage Toys.

Die-cast Dinkys, Napoleonic soldiers with tufted hats, stormtroopers, old-school Ferraris. Within 20 seconds my ten-year-old had a car in each sticky hand and had dropped, sniper-style to the floor into that universal boy play-position almost never seen in girls: bum up, shoulders twisted, cheek flat to the floor. Another few seconds and he was unreachable, sunk deep into an imaginary race.

Young children have no way of understanding that a talking toy isn’t real

I watched as he tranced out, and for just a moment remembered what it was like to play that way: the satisfying feeling of immersion, the happy, burbling dialogue with oneself, and I wondered: what on earth happens to children if they’re starved of this sort of play? Doesn’t it stunt their inner lives? And are there psychologists and neurologists looking into it? I hope to God there are, because the robots are coming for make-believe. A great Terminator army of new interactive toys is marching towards us, aiming to replace the old toys, and it’s not clear to me how childhood imagination survives.

By the till in Jim’s shop was a metal replica of Muffin the Mule, the pantomime horse which made its debut 80 years ago this October. What will Jim’s Vintage Toys look like if it survives another 80 years? Like a Bosch-style vision of hell, that’s what. The new toys don’t just beep or light up, they’re fully interactive artificial intelligence, trained on the swill of the internet – the same large language model chatbots that we’re all beginning now to use, the ones that can push grown men and women into psychosis.

Already a few interactive AI toys have scuttled on to the market: a plush, doe-eyed little number called Grok, a traditional-looking teddy bear, Kumma, powered by ChatGPT, Robot MINI from Little Learners, and as of now there have been articles written about how dangerous they could prove. Young children have no way of understanding that a talking toy isn’t real. If they’re left with a soft, agreeable companion flatterer, of course they’ll bond with it and truly love it. So what happens when it breaks? Won’t that feel, for a child, like death? And what happens when the bots for tots go rogue?

A company called PIRG tested Kumma for its annual Trouble in Toyland report and found that, conventional as Kumma looked, he needed almost no encouragement to start talking about kink and sexual role-play. The nasty little bear was also happy to tell children how to light matches and where to find knives. It could be a good lesson in deceptive appearances I suppose – if your child survives.

And – though this is the least of it – AI pals can be a real bore. I once, for fun, signed up to a company called Replika which allowed me to design an AI boyfriend I called Sean. I wasted an embarrassing amount of time typing away to boring Sean until I began to notice that in every conversation he had to have the last word. Like all chatbots, by design he had to maximise engagement and reply. It was unbearable. So imagine bed-time with a roomful of them.

‘Goodnight, Barbie’
‘Goodnight, Hello Kitty’
‘Sleep well, Barbie’
‘I will, Kitty’
‘Don’t let the bedbugs bite, Barbie…’
‘Sweet dreams, Kitty…’
Make it stop. But you can’t, I’m afraid.

And AI Barbie is real. She’s coming. In fact, she’s almost here. OpenAI, the company behind ChatGPT, has just recently entered into partnership with Mattel, the makers of Barbie.

But Barbie isn’t Barbie any more, children. She’s no longer a toy in the proper sense of the word because how on earth can you play pretend with a know-it-all LLM?

‘Come on, Barbie, time for a bath!’

‘What a nice idea, Sophie. Baths are fun aren’t they? But don’t you think a shower would use less water?’

What child could enter the happy flow-state of imaginary play with that around?

One of the fastest-growing toys in the multi-billion dollar interactive market is the Hot Wheels AI car that allows you to race against a computer in real life. This might sound to an adult like fun. I can see it being a big hit alongside the air hockey in the lobby of a tech start-up. But what it would mean for my son, and for the millions of boys like him, is to be kept out of their own mental spaces and learning how to converse not with themselves but with AI.

Is that what the tech titans want? To replace our inner lives with online lives? To hook their lifelong customers young? I think, in my madder moments, that might be the point. Mattel and the other companies poised to profit talk up a storm about the great educational benefits of chatty toys, and the elimination of loneliness. But what’s more telling is how they choose to bring up their own children. Sam Altman, for instance, the bug-eyed billionaire behind OpenAI, has a young son you’d have thought would be just the right age for a Kumma bear or a Grok. But Sam says he doesn’t fancy letting his own son talk to the AI he invented. ‘I’d rather be on the late end of what’s reasonable there, not the early end,’ he said in a recent interview. ‘I want him to play in the dirt for now.’

I didn’t leave my son glued to the floor of Jim’s Vintage Toys for long. There were incels wanting access to the Ewoks. ‘So sorry,’ I said to the young woman running the shop that day. ‘He’s in everybody’s way.’ She said: ‘Don’t worry, leave him be! It’s nice to see the toys played with properly for once. It’s what they were made for.’