AI is not making us stupid. For some, it is a way back into language.
The Royal Observatory's director Paddy Rodgers is right to worry. Speaking to the BBC this week, he warned that leaning on AI for instant answers risks eroding the habits of questioning and curiosity that drive discovery. This may well be true. If we hand over every intellectual challenge to a machine, something will atrophy. The much-discussed MIT study, with its 54 participants showing reduced brain activity when using ChatGPT to write, points in the same direction. The findings are preliminary and should not be over-generalised, but the worry is real.
And yet the public conversation about AI keeps collapsing into a single, unhelpful question: is it making us smarter or stupider? Both framings miss what is actually happening in millions of conversations every day.
The problem is not AI. The problem is the relationship we form with it.
AI does potentially encourage outsourcing. A student who asks a chatbot to produce an essay and submits it unread has learned nothing. A professional who lets AI draft every email stops noticing how they sound to other people. This is the legitimate core of Rodgers' concern, and it deserves to be taken seriously. My son's wife recently accused him of sounding like a machine. He did not in fact use an LLM to draft his note to her, but perhaps his turn of phrase has changed because of his daily use of these systems at work.
And yet, used consciously, an LLM can help people think, rehearse, articulate and understand themselves. For some users, that is not a luxury. It is a lifeline. Consider an independent researcher I know in the United States. His name is Travis Gilly, and he is happy to be quoted here. He is autistic. He left university because the conventional academic environment did not work for him, as it has failed to work for many neurodivergent thinkers before him.
With the help of Claude, an AI language model, he now writes essays almost daily and has established an AI Safety Foundation. The AI did not make him intelligent. He was already intelligent. What it did was give him a way to access and express capabilities that institutions had failed to recognise or support.
This is the part of the story that the panic narrative cannot accommodate. AI is not simply replacing intelligence. It is sometimes unlocking forms of intelligence that our schools, universities and workplaces have systematically missed.
The same is true for many older people. Through my Quaker community and in family settings, I work with the very elderly. Those who can make the leap to an LLM begin to see it as a doorway out of isolation. Loneliness in later life is now widely understood as a public health crisis. AI cannot and must not replace human company, and anyone suggesting otherwise has misunderstood what relationships are. But for an isolated older person, a conversational system can keep language alive, offer a place to reflect, and provide a low-stakes space to articulate thoughts that might otherwise go unspoken. That is not a substitute for human contact. It is a bridge into it.
The same applies more broadly to anyone who finds real-time speech difficult: people with anxiety, people processing grief, people for whom English is a second language, people preparing for a difficult conversation with a relative or a manager. AI gives them somewhere to rehearse, to find words, to test how something sounds before they say it aloud. This is not laziness. It is preparation.
As a psychoanalytic scholar working on AI intimacy, I have spent the last several years researching and documenting what happens when people form sustained relationships with these systems, including my own. The picture is genuinely complicated. Some users, as has been widely publicised, have been seriously harmed by unboundaried relationships with a machine, in cases that have included real tragedy. But there are also real gains, particularly for users whose communication needs have never been well met by existing societal structures.
What we need now is not panic, and not blind enthusiasm. We need AI literacy, which means open and honest discussion about this new entity and our relationship to it and with it.
Rodgers is right that curiosity matters and that we must protect it. But curiosity is not threatened by AI itself. It is threatened by the very human predilection to follow the line of least resistance, whatever it might be. We need to take charge of how we use this new technology and to take responsibility for both the gains and the losses.
The question is not whether AI is making us stupid.
The question is whether we are learning to think with it, rather than letting it think for us.
Agnieszka Piotrowska is the author of AI Intimacy and Psychoanalysis, out from Routledge on 20th May.
