Have you ever stopped to think about how AI might be reshaping the way we communicate—not just with machines, but with each other? It’s a question that’s been nagging at me lately, especially as I’ve watched the rise of large language models like ChatGPT. What’s particularly fascinating is how these models, despite their sophistication, are trained on a skewed slice of human language. They’re fed written texts—books, social media posts, scripts from movies—but they have minimal exposure to the unscripted, spontaneous conversations that make up the bulk of our daily interactions. This, I believe, is where the real danger lies.
The Silent Erosion of Natural Speech
One thing that immediately stands out is how AI’s limited training data could subtly erode the richness of our language. Personally, I think this isn’t just about losing a few words or phrases; it’s about losing the emotional depth and nuance that make human communication so unique. Take, for example, the way AI tends to produce smooth, polished text with a narrow vocabulary and sentence structure. A recent study from the University of Coruña highlighted this, showing that machine-generated language often lacks the meanders, interruptions, and leaps of logic that convey emotion. What many people don’t realize is that these imperfections are what make us human. They’re the cracks where our personalities shine through.
But here’s where it gets interesting: as we interact more with AI-generated text, we might start adopting these patterns ourselves. If you take a step back and think about it, this could lead to a homogenization of language, where our speech becomes more formulaic and less expressive. It’s like we’re trading in our colorful, imperfect human voices for something more robotic. And this isn’t just speculation—there’s already evidence of this happening. A 2022 study found that children in households using voice assistants like Siri and Alexa began speaking more curtly, issuing commands instead of asking politely. This raises a deeper question: are we teaching ourselves to communicate like machines, even as machines learn to mimic us?
The Feedback Loop of Inauthenticity
What this really suggests is that we’re caught in a feedback loop. AI models are trained on human-generated text, but as more of that text is produced by AI itself, the models start imitating their own inhuman patterns. And we, in turn, start imitating them. It’s a linguistic echo chamber, and it’s hard not to feel a bit uneasy about where it’s heading. From my perspective, this isn’t just about language—it’s about identity. How we speak shapes how we think, and if our speech becomes more rigid and less expressive, what does that mean for our ability to connect with others?
A detail that I find especially interesting is how AI’s lack of exposure to natural, unscripted speech affects its ability to emulate real conversations. For instance, when you tell ChatGPT, ‘I hate Beth!’ it responds with a formulaic three-part reply that feels utterly inhuman. No one talks like that in real life—at least not yet. But here’s the thing: the more we interact with these formulas, the more we might start accepting them as normal. It’s like we’re outsourcing our emotional intelligence to machines, and that’s a slippery slope.
The Biases We Don’t See
Another angle that’s often overlooked is how AI’s training data can introduce subtle biases into our thinking. Large language models are trained on written texts, which means they’re exposed to the worst of our online behavior—the flame wars, the toxic comments, the exaggerated opinions. What this really suggests is that AI might be learning to mirror our most divisive and aggressive tendencies, even as it tries to avoid them. This isn’t just about language; it’s about how we perceive the world. If AI amplifies our biases, it could make us less open to new ideas and more entrenched in our initial impulses.
Personally, I think this is one of the most troubling aspects of AI’s influence on language. As a teacher, I’ve seen students turn to generative AI for help with assignments, only to produce work that feels confident but lacks critical thinking. What many people don’t realize is that writing or speaking our thoughts is often how we clarify them. AI doesn’t help us think—it just regurgitates our half-formed ideas in a polished package. This, I believe, is a missed opportunity. Instead of deepening our understanding, AI might be encouraging us to settle for surface-level certainty.
The Way Forward: Reclaiming Our Humanity
So, where do we go from here? In my opinion, the solution isn’t to abandon AI—it’s to rethink how we train it. If we’re going to create models that truly understand and reflect human language, we need to include the full spectrum of how we communicate. That means finding ways to incorporate natural, unscripted speech into training data, even if it’s challenging. What makes this particularly fascinating is that it’s not just a technical problem—it’s a cultural one. We need to decide what kind of language, and by extension, what kind of society, we want to build.
If you take a step back and think about it, this is a pivotal moment. AI has the potential to either enrich our language or strip it of its humanity. The choice, I believe, is ours. But to make the right decision, we need to start paying attention to the subtle ways AI is already shaping how we speak and think. Because if we don’t, we might wake up one day to find that we’ve lost something irreplaceable—the messy, beautiful, uniquely human way we connect with each other.
