Machines have learnt to speak like us. Now let’s start speaking like them

For years, we have had a fairly straightforward concern: will artificial intelligence be able to speak like a human being? The answer, by now, is yes – quite often. The new problem is more amusing: will human beings be able to avoid speaking like artificial intelligence?

The Wall Street Journal has reported on professionals who are beginning to recognise in their own language mannerisms borrowed from chatbots: overly symmetrical sentences, flawless constructions, prefabricated enthusiasm, a certain penchant for the ‘it’s not about this, but about that’ line, and the habit of explaining even what nobody had asked to be explained. Having trained the models on billions of our words, we have now begun phase two: the models are training us.

This is the first useful lesson of the AI era: correctness is not a personality trait. ChatGPT can help us spot a typo, organise a line of reasoning, or find a clearer way of expressing an idea. The issue arises when we also eliminate the rough edges, the quirks, the favourite words, the questionable metaphors and the overly long sentences – in other words, precisely those imperfections through which we normally recognise who is speaking. If we all become clearer, more organised, more balanced and more ‘professional’, we may find that we have also become more interchangeable.

Lesson two: in the age of AI, it is no longer enough simply to be original; you must also be able to prove it. That is the most paradoxical part. A writer excluded from an award because he was suspected of having relied too heavily on artificial intelligence; a university that allows lecturers to run students’ essays through specialised detection software; publishers and lecturers who no longer ask themselves merely ‘is it well written?’, but also ‘who actually wrote it?’.

Culture, which for centuries has rewarded the ability to erase the traces of one’s work, is suddenly beginning to demand receipts. Keeping drafts, versions, notes and revision histories could become the literary equivalent of a receipt. Before long, writers will no longer simply hand over a manuscript to their publisher. They will hand over the manuscript, three notebooks, the Google Docs revision history and, if possible, a photograph of themselves struggling with chapter nine.

It is absurd, yet it offers a serious lesson: authenticity becomes all the more valuable precisely when it becomes harder to verify.

Lesson three: it’s best to stop thinking of AI as a machine that produces answers and start using it as a machine that creates friction. The most interesting way to use it isn’t ‘write this for me’, but ‘tell me where I’m going wrong’, ‘find the best counter-argument’, ‘show me what I’m taking for granted’, ‘where does my text resemble all the others?’. In other words: do not ask the machine to take the interesting work off your hands; let it do the work that enables you to get there.

Lesson four: knowing how to do something without AI will become a form of literacy, not nostalgia. Not because it is morally superior to do a division by hand, write a page without assistance or remember a route without a sat-nav. But because only those who possess a skill can tell when the tool that amplifies it is getting it wrong. If you cannot write, you cannot recognise an empty phrase. If you don’t know the subject matter, you cannot distinguish an elegant summary from nonsense. If you don’t have an idea, even the best prompt in the world will give you a highly polished version of nothing.

And finally, the most important lesson: being human does not mean doing worse at what machines do best.

We must not defend typos, slowness, forgetfulness or a poorly filled-in Excel spreadsheet just to prove we’re still alive. We must defend what is of value precisely because it cannot be standardised: taste, responsibility, experience, the courage to be wrong, personal memory, and the ability to choose one option when ten alternatives are all reasonable.

Because the danger is not that AI will become indistinguishable from us. It is that we, in order to appear more efficient, voluntarily decide to become indistinguishable from it.