What I Didn't Say

Large language models have made language itself a subject of public fascination. We ask machines questions, watch them generate essays, write code, compose music, and increasingly wonder whether they truly understand what they are saying. Yet beneath the excitement lies a deeper question that predates artificial intelligence: what is the relationship between language and meaning? This poem emerged from that question. While modern AI systems are often described as predicting the next word, human communication has never been merely about words. We imply more than we state, interpret beyond what is written, and derive meaning from context, culture, experience, and imagination. Much of what we understand is never explicitly said. As I reflected on how both humans and machines engage with language, I found myself drawn to the tension between syntax and semantics, between literal words and intended meaning, between prediction and understanding. The poem explores whether imagination, interpretation, and meaning are uniquely human traits or whether machines are beginning to traverse that space as well. 'What I Didn't Say' is less a statement about artificial intelligence and more a meditation on language itself — the imperfect medium through which humans, and now machines, attempt to model and understand the world.

I converted this poem into a video with AI assistance and produced this. Hope you enjoy.

Somewhere between language and meaning,
both humans and machines try to imagine the world.
We speak in words, but often mean something more.
Perhaps intelligence begins there too.

If languages come alive,
know now, languages live through machines,
machines don’t forget
with long context
what languages would be?

If machines know languages,
Are human, flawed and all,
We say but not mean,
Make up words that haven’t been,
How would machines model it all?

Human creative mind,
Can’t rely on syntax to find,
Semantics lets us play,
You interpret what I didn’t say!

For machines its black or white,
Syntax is the highlight,
Outside fact when it creates,
We label that it hallucinates!

Billion size models are good,
Predicting next word as it should,
Context gets longer,
Reinforcements stronger,
Yet language was flawed beneath the hood!

Languages for machines exist,
experts trained to speak with it,
But to imagine, to create,
That barrier no longer fits!

Models that imagine words and what they see,
Simulating worlds of how things could be,
Details need not stack,
meanings kept abstract,
They predict and they learn, like you and me!

Semantics lets us play,
You interpret what I didn’t say,
Beyond the words we write,
Beyond the black and white,
Perhaps that’s where true meanings stay.