I know how AI will bring us down

On the smooth marble concourse by the exit doors at Heathrow Airport I met my first cleaning robot. It was purple, made by a company called Mitie and about waist-height – the size and shape of a park bin. It ran on wheels, dragging a grubby mop behind it, and it was polite. As my small son and I stumbled into its path, it backed off smoothly like a well-trained butler. I apologised to it instinctively, after which it appeared to follow us. My son said: ‘Mum, it likes us!’ Then, when we reached the door: ‘Mum, can we take it home?’ Then: ‘Mum, wait! I don’t think it wants to sweep any more. It needs a rest!’

Last month I wrote about deepfake technology and my newfound fear of AI. In that moment on the concourse, I realised just how swiftly and easily it could manipulate us. Humans, especially British humans, are compulsive anthropomorphisers. We just can’t help but imagine that inanimate objects have personality. That cleaning robot didn’t even have a face and my seven-year-old was willing to sweep the floor for it. If Mitie made a bot that grimaced and whimpered as it worked, or tutted in a sorrowful way over humanity’s inability to recycle, it would in an instant attract a volunteer fleet of Brits trailing after it, litter-picking for free.

Experts reassure us that human intelligence is (so far) more sophisticated than AI. We don’t just crunch data and extract correlations; we seek explanations and have imagination. But what if it’s via our human imaginations that AI does us in? All AI has to do is work out how to make us pity it and the nation will be putty in its robotic hands.