Yesterday I wrote about how AI might be making us all think the same way. Today I want to explore the opposite possibility: that AI might actually perfect how we think by perfecting how we speak to each other.

Language shapes thought more than we realize. The words available to us determine which ideas we can express, which emotions we can articulate, which concepts we can grasp. Even punctuation affects how we think—the structures we use to organize ideas shape the ideas themselves. If AI is reshaping human language—and it clearly is—the question isn't whether this is good or bad. The question is whether it's making our language more capable of carrying human meaning.

And I think it might be.

English as Accidental Esperanto

For the past few decades, English has been becoming the world's de facto universal language. Not because of any inherent superiority, but because of internet infrastructure, American tech dominance, and network effectsNetwork effects in language work like they do in technology: the more people use a language for communication, the more valuable it becomes for everyone to learn it. This creates a winner-takes-all dynamic that can marginalize other languages.. Most programming languages use English keywords. Most international business happens in English. Most of the internet's content is in English.

This wasn't planned—it just happened. English became the Esperanto that Esperanto never managed to be: the common tongue that lets humans communicate across linguistic boundaries.

But this dominance came at a cost. Non-English speakers had to learn English to participate fully in the digital world. English speakers got lazy about learning other languages. Rich linguistic traditions got marginalized online. The internet's early architecture essentially encoded English linguistic imperialism into its DNA.

LLMs are changing this dynamic in ways we're just beginning to understand.

The Great Translation

AI translation has crossed a threshold. Not perfect, but good enough that people can have genuine conversations across language barriers. Good enough that Korean novels get read in real-time by Spanish speakers. Good enough that Finnish programmers can contribute to open source projects documented in Portuguese.

class LinguisticDemocracy:
    def __init__(self):
        self.barriers_removed = []
        self.voices_amplified = []
        self.ideas_cross_pollinating = True
        
    def translate(self, thought, from_language, to_language):
        # Not just words—cultural context, emotional nuance
        translated_meaning = preserve_essence(thought)
        adapted_context = cultural_bridge(translated_meaning)
        return make_accessible(adapted_context, to_language)
        
    def amplify_voices(self):
        # Every language becomes globally accessible
        for language in HUMAN_LANGUAGES:
            if language.has_unique_concepts():
                make_globally_available(language.concepts)

This isn't just convenient—it's democratizingThis threshold isn't about technical accuracy—it's about preserving intent and emotional nuance across cultures.. Ideas that were trapped in specific linguistic communities can now cross boundaries. Concepts that only existed in particular languages can enrich global discourse.

Beyond Translation: Language Enhancement

But LLMs do something more interesting than translation. They help people express themselves better in their own languages.

Non-native English speakers can use ChatGPT to refine their writing—not to sound more "English," but to sound more like themselves. The AI doesn't homogenize their voice; it amplifies itThis is the key difference from traditional language instruction, which often erases cultural voice in favor of "correct" usage. AI can preserve the speaker's conceptual framework while helping them find precise expression within any language.. It helps them find the English words that carry their intended meaning most precisely.

Native speakers benefit too. How many times have you had a thought you couldn't quite articulate, then asked an LLM to help you find the words? The AI doesn't give you foreign thoughts—it helps you discover what you were already thinking.

I'm using AI to help me articulate these very thoughts you're reading right now.

LLMs have access to essentially the entire documented vocabulary of human language. They know archaic words, technical terms, poetic constructions, regional dialects. When you're searching for exactly the right word, they can offer options you didn't know existed.

This isn't making language more uniform—it's making it more precise. More nuanced. More capable of carrying complex meaningThe opposite of what we feared in The Mirror—instead of averaging out linguistic diversity, LLMs might be preserving and amplifying it by making obscure vocabulary accessible to anyone who needs it..

Instead of everyone converging on the most common words, people can discover the exact words their thoughts require. The result could be more expressive, not less.

Cultural Cross-Pollination

Here's where it gets really interesting: LLMs trained on multilingual data can introduce concepts from one language into another. They know that Japanese has mono no aware (the bittersweet awareness of impermanence), that Portuguese has saudade (longing for something absent), that German has verschlimmbessern (to make something worse by trying to improve it)These "untranslatable" words aren't actually untranslatable—they represent concepts that one culture developed precise vocabulary for while others expressed the same ideas through longer phrases or circumlocution. AI can make these conceptual tools universally accessible..

These concepts can now migrate across language boundaries not as foreign curiosities, but as useful tools for thought. English speakers can borrow hygge from Danish not just as a lifestyle trend, but as a genuine enhancement to their emotional vocabulary.

Will they? Only time will tell.

AI is also democratizing sophisticated grammar. Complex sentence structures that required years of study to master can now be deployed by anyone. Rhetorical techniques from classical rhetoric are becoming accessible tools rather than academic artifacts.

This doesn't mean everyone will write like Victorian novelists. It means everyone can access the full range of their language's expressive potential when they need it.

Preserving the Endangered

One of the most profound possibilities: LLMs could help preserve and revitalize endangered languages. By learning from limited texts and native speaker interactions, AI could become a bridge between disappearing linguistic traditions and new generations of speakersThis reverses the historical pattern where technological advancement often accelerated language loss. Instead of forcing assimilation to dominant languages, AI could make preserving minority languages economically and practically viable..

Indigenous languages that were being lost could find new life in digital spaces. Cultural knowledge encoded in specific linguistic structures could be preserved and transmitted.

The Optimistic Scenario

Here's what linguistic evolution through AI could look like:

  • Precision over simplification: Instead of everyone using the same basic vocabulary, people discover exactly the words their thoughts require.
  • Cultural exchange: Concepts from different languages enrich global discourse rather than being lost to English dominance.
  • Accessibility: Complex grammatical structures and sophisticated rhetoric become available to anyone, not just linguistic elites.
  • Preservation: Endangered languages and dialects get documented, preserved, and potentially revitalized.
  • Enhancement: Everyone becomes more articulate in their own language, better able to express their actual thoughts and feelings.

Instead of convergence toward linguistic mediocrity, we might see an explosion of expressive capability. Writers could seamlessly incorporate techniques from multiple traditions. Speakers could deploy exactly the rhetorical tools their ideas require. Poets could draw from the entire heritage of human language.

As translation becomes effortless, English's role as universal intermediary might diminish. Korean speakers could communicate directly with Arabic speakers without English as the bridge language. Scientific papers could be written in researchers' native languages and instantly accessible globally. This could restore linguistic balance to the internet while maintaining global connectivity.

The Paradox Resolution

This resolves the apparent paradox from The Mirror: AI might homogenize how we think while diversifying what we can think about. The cognitive patterns converge, but the available vocabulary expands. The processing mechanisms become similar, but the expressive range increases.

We all learn to think in prompts, but those prompts can deploy concepts from every human language that ever existed.

What We're Potentially Gaining

  • Linguistic precision: The exact words for our exact thoughts.
  • Cultural concepts: Emotional and intellectual tools from every tradition.
  • Expressive range: Access to sophisticated grammatical structures.
  • Global connectivity: Communication without linguistic barriers.
  • Preserved diversity: Endangered languages and dialects maintained digitally.
  • Enhanced articulation: Everyone becomes more eloquent in their native tongue.

The Path Forward

This optimistic scenario isn't guaranteed. It requires:

  • LLMs trained on truly diverse linguistic data, not just English-dominant corpora.
  • Translation systems that preserve cultural context, not just literal meaning.
  • AI tools designed to enhance rather than replace human expression.
  • Platforms that reward linguistic creativity rather than algorithmic optimization.
  • Educational approaches that use AI to expand rather than limit vocabulary.

But if we build it right, AI could give us the best of both worlds: global connectivity and linguistic diversity, efficiency and expressiveness, accessibility and sophistication.

The mirror might be teaching us to look the same, but the translator could be teaching us to speak like ourselves—only better.

On This Site

External Resources

  • Through the Language Glass by Guy Deutscher - How language shapes thought
  • The Language Instinct by Steven Pinker - The cognitive basis of human language
  • Babel: Around the World in Twenty Languages by Gaston Dorren - Linguistic diversity and evolution