Could the AI have a fault? The answer affords us a delightful quarter of an hour of smug satisfaction: it doesn’t have hands

As things stand at the moment, robots can’t dig: even to make a hole in the ground and plant a basil seedling the right way round, you need skill – you need human hands
23 JUL 26
Translated by AI
Image of Could the AI have a fault? The answer affords us a delightful quarter of an hour of smug satisfaction: it doesn’t have hands

Photo by Cash Macanaya on Unsplash

I must confess that almost all my work emails are written by my AI; I’ve called it ‘The Beast’, and it replies to most clients and spammers. By now, my reliance on this tool has become second nature and is no longer a viable option: I simply cannot be bothered to write informative paragraphs on my own any more; I just cannot do it. I’ve called it ‘The Beast’ because my AI always obeys, works quickly – much faster than trainees – and never gets the emails wrong. Even if I give it convoluted and confusing instructions, ‘The Beast’ is intuitive and understands what needs to be understood. It’s not stupid at all, if you give it the right orders: it streamlines, summarises, strikes a firm yet polite tone; if you ask it to, it becomes optimistic or mildly pessimistic about the best course of action to resolve the employment law issue. It’s also a psychologist down to the last detail: it can be tough or gentle on command, so I can relax and not have to think about it.
In this land of plenty, however, something is gnawing away at my smile: is it possible that La Bestia might become better than me? Possible – in fact, certain. And so I’ve recently become aware of a conditioned reflex of which I’m rather ashamed: given that everything around me is now a torrent of beautiful, orderly (and lifeless) writing, I force myself, when I write, to write with a human twist – which must necessarily be original – to improvise a new syntax, to make Gadda’s bones tremble by imitating those styles. I use dialect, my grandfather, my own experiences, Luzzano (Bn)—anything to stay a few metres ahead of the Beast. How long will it last, I wonder? How long will it take them to carry out that fatal update—the one that creates a total, indistinguishable replica of anyone, from Roth to Berselli? How do I plunge thirty-three knives into the Beast’s back, or at the very least, how do I manage to keep it on a leash? That is the question. Do these synthetic intelligences have a weak spot? It seems so: their hands. And if we were to take a liking to metaphors, this one would be exceptional.
Hands are the masterpiece of the human ape (us). Sensitive, flexible and capable of extremely precise touch. If that precise touch is practised, it can also become swift and automatic. Tying your shoelaces. Or moving extremely fragile objects. Engineers are currently going mad: machines are making progress, but the margins of error are minuscule. It takes the ‘Beast’ several minutes to fold a T-shirt, and loading a dishwasher or making spaghetti with clams is out of the question – that’s a distant future. We read in the newspapers that robotic hands represent the most complex challenge in robotics. In China, they’re working day and night on these humanoids, which perform pirouettes and other pointless little tricks but have hands like scissors or shortcrust pastry. In short, if a robot can move but cannot carry out most everyday tasks, it’s useless; at best, we’ll have parcel-sorting robots, nothing more. Elon Musk has described the hands as the component responsible for most of the engineering challenges facing Tesla’s Optimus robot. According to Zhou Yong, founder of LinkerBot, as quoted by The Guardian, building an artificial hand can be “a hundred times more difficult” than constructing the rest of a humanoid. Mobility, sensors and high-precision components must all be packed into a very small space.
It is the fastest-growing sector: according to Chinese media, the domestic market for articulated robotic hands is reported to have exceeded 50 billion yuan, up from 13 billion in 2024. The difficulty appears to lie in the software. Text and images alone are not enough to train a robot: three-dimensional data on movement, pressure, contact and the texture of objects are required. There are two fundamental problems: accurately replicating human movement and distinguishing between what requires one type of touch and what requires another. As things stand at the moment, robots couldn’t even hoe a patch of soil: even to dig a hole in the ground and plant a basil seedling in it the right way round, you need artistry – you need human hands. Hands. Which are, after all, instinctive; sometimes they react before our thoughts even catch up. Everything that makes life joyful is done with our hands; love begins with our hands, and it is always our hands that are the first to hold back the tears. The Beast, for now, cannot manage it. Let us enjoy this quarter of an hour of superiority.