November 24, 2025 05:38 PM IST
First published on: Nov 24, 2025 at 07:10 AM IST
There is a lovely absurdity to how we speak about Large Language Models, as if we are sculpting a human mind minus the clutter that makes a mind human. We want intuition without confusion, memory without distortion, brilliance without doubt. It is like opening a jazz club, banning improvisation and calling it progress.
For years, we trained machines to sound like us. Now we shrink back when they begin sounding too much like us. The moment they imagine, we label it a malfunction. The moment they offer a metaphor, we reach for hazard signs. We built a mirror and panicked when the reflection began thinking for itself.
Humans were never monuments of precision. We survive not by accuracy but by instinct. The first person who saw lightning and imagined God probably lived longer than the one who waited for data validation. Civilisation itself rests on elegant speculation. Yet here we are, swirling the glass like connoisseurs of truth, frowning when a model “hallucinates”.
Claude Lévi-Strauss would have chuckled at our anxiety. Myth, he said, was the engine that cultures built to manage contradiction. A story was a way of stitching chaos into coherence. When an LLM fabricates a quote, it is doing what humans have always done: Smoothing the wrinkles of uncertainty into a narrative we can bear. Perhaps an LLM is just the newest apprentice storyteller.
What is meaning if not a beautifully organised hallucination? Jacques Derrida reminded us that language is forever slipping away from itself. No word stands still. Every sentence gestures towards another . So when a machine fills gaps or misremembers with confidence, it is not betraying intelligence. It is participating in the oldest grammar of being human. Societies behave the same way. Flags are hallucinations dyed into fabric. Money is a hallucination printed on special paper. We survive on shared fictions.
Somewhere in history, we decided that accuracy was sacred. We mistook truth for housekeeping, error for moral failure. Yet, perfection has never been the purpose of thought. Our fear of AI hallucination reveals a deeper issue: That intelligence without uncertainty feels mechanical, and certainty without imagination feels sterile.
Ernst Kapp wrote that tools extend our bodies. The hammer extends the fist. The wheel extends the foot. By that logic, the LLM is an extension of our tongue, a borrowed mouth holding the sediment of everything we have ever said. Yet, we scold it for speaking too freely.
There is something strangely touching about a model that misquotes Friedrich Nietzsche with conviction. It feels recognisably human. It has read everything, from scripture to gossip, and is now whispering its way through the muddle. When humans invent, we call it imagination. When poets do it, we call it art. When marketers do it, storytelling. When machines do it, we call it hallucination. Perhaps originality is tolerated only when it stays carbon-based.
Friedrich Dessauer warned that technology without imagination becomes a blind instrument. The reverse is true as well. Imagination without room for error becomes propaganda. If we force these models into factual obedience, we will end up with machines that are flawless but lifeless, precise but pointless — the spiritual equivalent of a corporate mission statement.
The miracle of LLMs is not their accuracy. It is their music. When a model hallucinates, it is doing what we do when truth alone feels insufficient: Reaching for metaphor to make the world slightly more bearable. Perhaps the LLM is our echo. An accidental therapist repeating our own contradictions back to us.
We call it artificial intelligence. But what if it is the most human thing we have ever built? It dreams, stumbles, overgeneralises, contradicts itself, and occasionally produces poetry. Maybe the machine is not hallucinating. Maybe it is improvising. Maybe the day we build a model that never hallucinates will be the day imagination quietly retires. Because a mind without a dream is not intelligence. It is only grammar with electricity.
The writer is an advisory professional
