You look like chatgpt | The penis

Join any zoom call, enter any conference room or watch any YouTube video and listen carefully. Beyond the content and inside linguistic patterns, you will find the rampant uniformity of the voice of AI. Words like “prowess” and “tapestry”, which are favored by Chatgpt, slip into our vocabulary, while words like “meeting”, “unarth” and “nuance”, the words less favored by Chatgpt, have decreased. The researchers already document changes in the way we speak and communicate following Chatgpt – and they see this linguistic influence accelerate in something much greater.

During the 18 months after the release of Chatgpt, speakers used words as “meticle”, “delve”, “kingdom” and “adept” up to 51% more frequently than in the previous three years, according to researchers from the Max Planck Institute for Human Development, who analyzed nearly 280,000 YouTube videos academic channels. The researchers excluded other possible points of change before the release of Chatgpt and confirmed these words align with those that the model favors, as established in a Previous study comparing 10,000 human edited texts and AI. The speakers do not realize that their language changes. This is exactly the point.

A word, in particular, stood out for researchers as a kind of linguistic watermark. “Delve” has become an academic shiboboleth, a sign of neon in the middle of each flashing conversation Chatgpt was there. “We internal this virtual vocabulary in daily communication”, explains Hiromu Yakura, principal author of the study and postdoctoral researcher at the Max Planck Institute of Human Development.

But it is not only that we adopt the language of AI – this is the way we start to ring. Even if current studies focus mainly on vocabulary, researchers suspect that the influence of AI also begins to appear in tone as well – in the form of a longer and more structured word and an emotional expression in mute. Like Levin Brinkmann, researcher at the Max Planck Institute for Human Development and co-author of the study, puts it, “” Delve “is only the tip of the iceberg”.

The AI ​​appears most obviously in functions such as intelligent responses, automatic correction and spelling verification. Cornell research Look at our use of intelligent responses in cats, noting that the use of intelligent responses increases global cooperation and proximity feelings between participants, because users end up selecting a more positive emotional language. But if people thought their partner used AI in interaction, they judged their partner as less collaborative and more demanding. Above all, it was not a real use of the AI ​​that extinguished them – it is suspicion. We form perceptions based on linguistic indices, and it is really the linguistic properties that stimulate these impressions, explains Malta Jung, associate professor of information sciences at Cornell University and co-author of the study.

This paradox – AI improving communication while promoting suspicions – underlines a deeper loss of confidence, according to Mor Naaman, professor of information sciences at Cornell Tech. He identified three levels of human signals that we have lost in the adoption of AI in our communication. The first level is that of the basic signals of humanity, of clues that speak of our authenticity as a human being as moments of vulnerability or personal rituals, who say to others: “It’s me, I am human.” The second level consists of signals of attention and effort which prove “I cared enough to write that myself”. And the third level is the capacity signals that show our sense of humor, our skill and our real self for others. This is the difference between sending sms to someone, “I’m sorry you are upset” against “hey sorry I panicked at dinner, I should probably not have jumped therapy this week.” We seem flat; the other seems human.

For Naaman, finding how to bring back and raising these signals is the way to follow in communication mediated by AI, because AI does not only change the language – but what we think. “Even on dating sites, what does it mean to be funny on your profile or in your cat where we know that AI can be funny for you?” Naaman asks. The loss of the agency that begins in our speech and goes up in our reflection, in particular, is what is worried. “Instead of articulating our own thoughts, we artigulate everything that helps us to articulate … We become more convinced.” Without these signals, warns Naaman, we will only trust face -to -face communication – not even video calls.

The problem of confidence is made up when you consider that AI is quietly established which may seem “legitimate” in the first place. University of California, search for Berkeley have found that AI’s responses often contained stereotypes or inaccurate approximations when they were invited to use dialects other than standard American English. Examples of this include the Chatppt repeating the invitation to the non-standard-American-English user due to the lack of understanding and considerably exaggerating the input dialect. A Singaporean English respondent commented“The super exaggerated song in one of the answers was slightly worthy of Digne.” The study revealed that AI does not only prefer standard American English, it actively flattens other dialects so as to lower their speakers.

This system perpetuates inaccuracies not only on communities but also on what is “correct” English. The stakes therefore do not consist of preserving linguistic diversity – it is a question of protecting imperfections which really reinforce confidence. When everyone around us begins to appear “correct”, we lose verbal stumbling trees, regional idioms and sentences outside Kilt which signal vulnerability, authenticity and personality.

We are approaching a point of division, where the impacts of AI on the way in which we speak and write a movement between the poles of normalization, such as the models of professional email or formal presentations, and an authentic expression in personal and emotional spaces. Between these posts, there are three basic tensions at stake. The first counterposter signals, such as academics avoiding “diving” and people who actively try not to resemble AI, suggests that we can self -regulate against homogenization. The AI ​​systems themselves will probably become more expressive and personalized over time, potentially reducing the current Vocal problem of AI. And the deepest risk of all, as Naaman pointed out, is not linguistic uniformity but losing conscious control of our own thought and our expression.

The future is not predetermined between homogenization and hyperpersonalization: it depends on whether we will be aware of the participants in this change. We see the first signs that people repel when the influence of AI becomes too obvious, while technology can evolve to better reflect human diversity rather than flatten it. It is not a question of knowing whether the AI ​​will continue to shape how we speak – because it will do – but if we will actively choose to preserve space for verbal whims and emotional disorder that make communication in a recognition, irreplacebly human.