The Language Game We Can No Longer Keep Up With

Language has always been a human game, until now, when the machines started playing too fast for us to keep up.

The Game Board Is Changing

Philosopher Ludwig Wittgenstein described language as a “game,” a system of rules and practices we all instinctively navigate. For millennia, humans were the only players. Today, Large Language Models like ChatGPT have entered this game, operating at speeds and scales that fundamentally alter the playing field itself.

AI pioneer Yann LeCun recently argued that his field has been “led astray” by LLMs. He’s right to be concerned. The problem isn’t that LLMs are sophisticated. It’s that they’ve transformed language from a human-scale practice into what philosopher Timothy Morton calls a “hyperobject”: something so vast and nonlocal that we can only experience fragments of it.

LLMs don’t just play the language game faster than humans. They are rewriting the game board, turning language into infrastructure that operates at machine scale while humans remain metabolically constrained, cognitively limited players.

The Scale Mismatch: When Language Outruns the Mind

Humans evolved for small-scale linguistic interaction: face-to-face conversations, stories shared around fires, slow-paced debates. Our brains are predictive, narrative-making machines designed for embodied communication. We think at roughly 40 bits per second. We need sleep, food, and time to process meaning.

LLMs process millions of tokens per second. They generate coherent language endlessly without fatigue, distraction, or metabolic limits. They are tireless linguistic juggernauts bound only by code and data.

This creates a radical asymmetry. When you ask ChatGPT a question, you’re not engaging with a peer in the language game. You’re interfacing with a system operating at scales your cognition cannot fully metabolize. The result is a phenomenon called Cognitive Hollowing.

Cognitive Hollowing: What Happens When Language Outruns the Mind

Cognitive hollowing occurs when humans shift from being generators of meaning to interfaces for machine-scale language. It describes the erosion of human cognitive engagement, creativity, reasoning stamina, and nuanced meaning-making due to overexposure and dependence on AI-generated text.

This happens in five interconnected ways:

Memory externalization. When everything can be recalled instantly by an AI, we forget how to remember. Not just facts, but the associative networks that give facts meaning.

Reasoning outsourcing. When machines handle complex logic, our mental stamina weakens. The “thinking muscles” atrophy from disuse.

Imagination delegation. Letting AI generate ideas leads to creative atrophy. We begin with AI suggestions rather than internal synthesis.

Attentional collapse. The flood of machine-generated content fragments our focus and flattens our capacity for sustained engagement.

Nuance erosion. AI linguistic smoothing removes the rough edges: the pauses, ambiguities, and emotional textures that hold cultural and personal depth.

Here’s what this looks like in practice: You’re writing an email and let autocomplete finish your sentences. You’re researching a topic and accept ChatGPT’s first answer without verification. You’re brainstorming and scroll through AI-generated options instead of sitting with uncertainty. Each instance seems harmless. But cognitive hollowing is cumulative. Over time, your tolerance for ambiguity shrinks. Your stories grow thinner. Your identity becomes more fragile because it’s less internally authored.

Lessons from History: Outsourcing Cognition Over Time

Cognitive hollowing is not new, but AI accelerates and deepens it on an unprecedented scale.

Calculators, once essential mental math skills fade as handheld devices do the work effortlessly
Remembering phone numbers, digital contacts replaced memorization, changing how we encode personal relationships
Internet search, access to instant global knowledge reshaped memory and knowledge synthesis, encouraging quick retrieval over deep comprehension

Each technological advance extended intelligence outward but at cognitive cost. LLMs extend this outsourcing into language itself, our core reality space.

The Crucial Difference: Prediction vs. Perception

When ChatGPT invents a fake legal citation, we call it a “hallucination,” but that’s anthropomorphizing prediction. Machines don’t hallucinate; they autocomplete at scale.

Human hallucinations arise from perception without external stimulus. They’re subjective experiences tied to emotion and survival. When you see patterns in static or hear voices in white noise, your brain is doing what it evolved to do: generate meaning from ambiguity.

LLM “hallucinations” are different. When an AI fabricates information, it’s computing the next most probable token sequence based on statistical patterns in training data. There’s no knowing or not-knowing, no belief or deception. Just probability distributions over language without grounding in reality.

This difference is profound. Machines don’t hallucinate: they predict. Humans hallucinate: we perceive. Both produce errors, but only humans experience being wrong. This asymmetry matters because it reveals that LLMs are not conversational partners. They are language-generation infrastructure we mistake for participants in dialogue.

The Attention Economy as Linguistic Infrastructure

The cognitive hollowing effect intensifies when you realize LLMs don’t just answer questions. They shape what you watch, buy, scroll, and think about.

Behind every recommendation on streaming platforms, online stores, and social media feeds are neural networks and machine learning algorithms tirelessly predicting and nudging preferences. These algorithms leverage linguistic cues, emotional triggers, and social proof to maximize engagement and consumption. They generate the flood of content that saturates your attention before you ever consciously engage with AI tools.

This is algorithmic intermediation at scale. Your brain, evolved for slow embodied interaction, now faces a relentless stream of machine-optimized language designed to capture and monetize attention. The space for reflection shrinks. Original thought becomes harder to locate. Instead of active cognition, you begin to passively consume and respond.

Recent research confirms this. A 2025 study found that AI usage negatively predicts critical thinking, and that trust in AI amplifies this effect, unless users develop high AI literacy. Another study documented that AI tools encourage “cognitive offloading” that weakens independent reasoning over time.

The language game has become environmental. You’re not just playing with AI. You’re playing inside a linguistic ecology increasingly shaped by machine-scale processes you can barely perceive.

What Thoughtful Engagement Looks Like

The language game has irreversibly changed. The question isn’t whether to play alongside digital players, but how to preserve human cognitive resilience while doing so.

This requires new literacy: not just using AI tools, but understanding when to engage them, when to resist, and when to reclaim the slow, metabolically-constrained thinking that makes us human.

Practical cognitive resilience means:

Mindful attention management. Recognizing when you’re interfacing with machine-scale language and deliberately creating space for unmediated thought.

Reclaiming internal memory. Practicing recall, synthesis, and associative thinking rather than outsourcing everything to external systems.

Critical evaluation of machine output. Developing the habit of verification, not just acceptance. Treating AI suggestions as starting points, not conclusions.

Reconnecting language to embodied experience. Writing, speaking, and thinking in ways that preserve emotional texture and personal nuance: the elements AI smooths away.

Rather than fearing or rejecting AI, we must evolve to engage with these linguistic hyperobjects thoughtfully. This isn’t about nostalgia for pre-digital cognition. It’s about consciously choosing which cognitive functions to preserve, which to augment, and which to delegate.

The future belongs to those who can navigate machine-scale language without being hollowed out by it. That’s not a technical challenge. It’s a cognitive discipline. And it’s the most important literacy challenge of our time.