Most people assume AI research is all math, code, and endless datasets. They imagine whiteboards filled with equations and engineers arguing about model accuracy.
That part is real.
But there’s another side to my work that surprises people — I write poems about robots.
It started accidentally. One evening after hours of debugging a language model, I typed a strange prompt into it: “Write a poem about a robot learning loneliness.” I expected nonsense. Instead, the model generated something clumsy but oddly touching. The lines weren’t perfect, but there was a hint of something poetic hiding in the machine’s patterns.
Since then, poetry has become my side experiment in AI research.
Not because machines understand emotions — they don’t. At least not the way humans do. But they can mirror language in ways that make us rethink what creativity actually is.
In the lab during the day, I train models, tune parameters, and measure performance metrics. Everything is structured. Precision matters. But late at night, when the office quiets down and the GPUs keep humming, I open a blank document and start writing.
Sometimes I write with the model.
I’ll write a few lines about circuits dreaming or servers whispering to each other across the internet. Then I let the AI respond with its own lines. The result feels less like programming and more like collaboration — a strange conversation between logic and imagination.
Of course, the robot isn’t truly “feeling” anything. It’s predicting words based on patterns. Yet when those patterns align just right, the output can feel surprisingly human.
That’s the fascinating tension in AI.
We build systems based on mathematics, but they interact with one of the most human things we have: language.
Poetry helps me explore that boundary. It reminds me that technology isn’t only about efficiency or automation. It can also be a mirror reflecting our own creativity back at us.
When I publish these poems online, people often ask the same question: Did you write it, or did the AI?
The honest answer is both.
Because somewhere between code and metaphor, between algorithms and imagination, there’s a new kind of storytelling emerging.
And sometimes, a robot ends up writing the final line.