By day, my world is datasets, models, and long experiments that run while I sleep. I work with machines that learn patterns faster than I can explain them, tuning parameters, checking loss curves, arguing gently with results that refuse to behave. By night, when the lab goes quiet and the screens dim, I write poems—often about the same machines I study all day.
It started unintentionally. After months of training neural networks, I noticed how often we speak about them in metaphors. We say models “see,” “understand,” or “hallucinate.” These words are shortcuts, but they also reveal something human: our need to relate. Poetry became my way of slowing down and questioning those metaphors instead of blindly trusting them.
Research demands precision. A model either converges or it doesn’t. The benchmark is passed or failed. Poetry, on the other hand, allows ambiguity. A robot in a poem can hesitate, dream, or misunderstand—not as a bug, but as a reflection of us. Writing helps me notice the quiet parts of my work: the assumptions baked into data, the values hidden inside objective functions, the things we don’t measure because they’re inconvenient.
There’s a strange balance between code and verse. During the day, I reduce complexity. At night, I let it expand. One hour I’m debugging why an agent behaves unpredictably; the next, I’m writing about a machine that learned sadness from corrupted data. The poems aren’t explanations. They’re questions. What does intelligence look like when stripped of performance metrics? What do we project onto systems that cannot answer back?
Some colleagues find this amusing. Others find it unnecessary. But for me, it’s grounding. Research moves fast—faster models, bigger claims, louder headlines. Poetry pulls me back to patience. It reminds me that intelligence is not just about output, but about context, responsibility, and restraint.
I don’t believe machines will become poets anytime soon. But I do believe humans need poetry more than ever as we build increasingly powerful systems. Not to romanticize technology, but to stay aware of our role within it.
In the end, my poems are not really about robots. They’re about us—standing in front of our own creations, listening carefully, and wondering what kind of future we are quietly training into existence.