AI: when humans become suspects
As we teach machines to speak like us, we’re starting to speak like them
All it takes is one word. Just one word too many, too bland, too clichéd, too predictable to make everyone a suspect. Suspicion arises as surely and as quickly as emails are opened. Whether the message is too polished, the phrasing too smooth, or the heart of the matter takes us on a “dive,” doubt creeps in, imperceptibly, inexorably. Or take chatbots: they love adverbs so much they overuse them to the point of excess. Suspicion, once reserved for a certain elite who, through a misunderstanding, could pass themselves off as artists, has become widespread. Generative artificial intelligence (AI) solutions, true personal assistants, are also, and above all, taking up the pen. Skepticism now hangs over messages, whether sent by email or posted on social media, just as it once did over students’ exam papers.
Every craze has its stock phrases and mannerisms. Humanity has been built on imitation. In this regard, technology excels. Thus, it has unearthed the now-famous em dash, a punctuation mark no one had ever heard of during their school years. This long punctuation mark has become the symbol and telltale sign of AI overuse, a signature, in short.
— Who would dare use it anymore?
Prose has become concise. Adjectives drone on. “It is important to note that” doesn’t signal anything important. Nowadays, arguments are presented in bullet points. It must be said that PowerPoint had already primed people’s minds. Newspeak is no longer the exclusive domain of corporate meetings. It’s taking over dinner parties among friends eager to prioritize and make an impact, only to better take offense at it and showcase their resilience as the ultimate mark of sophistication. The mundane becomes shaky. And vice versa.
Anthropomorphism. The most unsettling thing isn’t really that the machine imitates us. It’s that we imitate it in return. Chatbots feed on words and regurgitate them. We make them our own to revel in them and, in a way, claim authorship. Chewed over and rehashed, ideas drown in their own copies. Anthropomorphism lies latent. Our digital comfort blankets become technological animals. Conversely, the extension of our brains, RAM in our hands, paves the way for technical externalization. Ernst Kapp, a German-American geographer and philosopher, had described this as early as the 19th century. Throughout history, human beings have distinguished themselves by their ability to make tools and transform them into prostheses. The hand becomes a hammer. Thought turns into an algorithm. Being and machine become one. But this remains shameful.
The use of machines in the creation of intellectual works, whether graphic or written, must indeed be acknowledged so as not to be rejected
Faced with the unspeakable, a defense is mounting. To exonerate ourselves, we must now compete in imperfections. Thus, grammatical errors in French become a mark of authenticity.
Let’s be careful not to be too polite to the point of losing our honesty. Must we therefore be incorrect to stand out?
Automatic messages, sent mechanically, are commonplace. The suggestion of the next word in instant messaging was merely the prelude to an increasingly delegated form of writing. Beyond that, the use of machines in the creation of intellectual works, whether graphic or written, must be explicitly acknowledged to avoid rejection. Yet no field is immune to this.
A recent study by the Institut Agro Rennes-Angers, the University of Sydney, and the Queensland University of Technology shows that scientific research is facing a surge in fake publications fueled by AI. Thus, “more than 250,000 cancer-related scientific articles may have been fabricated from scratch between 1999 and 2024,” as reported by The Conversation in an article by Baptiste Scancar and David Causeur - Nomen est omen: the name is the omen. Fictitious or bogus articles now account for 15% of annual scientific publications, compared with 1% thirty years ago.
Detection software has its own limitations when it comes to identifying these fabricated articles. For example, Sciences-et-Vie reports that “An AI claims that the 1776 U.S. Declaration of Independence was… generated by an AI.” Thomas Jefferson, a geek ahead of his time.
“Style is the man himself,” declared the Comte de Buffon in his acceptance speech at the Académie française on August 25, 1753; for him, “it can therefore neither be removed, nor transferred, nor altered; if it is elevated, noble, sublime, the Author will likewise be admired throughout the ages; for only truth is enduring, and even eternal.” Because language carries the soul of the one who wields it, to paraphrase a quote attributed to Cervantes, resorting to artificial devices without acknowledging it undermines the bond of trust forged with the reader. One must write truthfully.
This article was originally published in French in l’Opinion