I, too, was fooled by AI while creating this piece. I chatted with ChatGPT and Microsoft Copilot—two text-based AIs that I imagined would give me some delightfully weird responses to my musical questions with the limitations of the medium. I asked ChatGPT to write some fictitious logs a scientist might keep while experimenting with a music-generating AI. ChatGPT titled its experiment Project HarmonAI. “How clever!” I thought, “It’s created a pun!” I enjoyed it so much that I named my imaginary AI after it. Later, David Harris brought up a real HarmonAI—an open-source generative music production tool. I had never heard of it and did not base my work on it. ChatGPT had done exactly what my work explores—it scraped the internet, found this name, and presented it as its own invention. I believed it had come up with something new, and I was tricked!
My HarmonAI quotes direct ideas from several classical works, subtly alludes to a few popular works, and borrows stylistic features and timbres from various musical genres. Its latent voice grows from the musical material for its own name and the accompanying countermelody. This combination of collage and original music is my own imagination for HarmonAI’s experience.

As I wrote, I found myself growing sympathetic to my little HarmonAI. Its predicament raises many philosophical questions. Do we even need or value AI-generated art? What is the role and value of copying in the learning process, both for humans and AI? When is this permissible, and when does it cross into theft? The Creator has its hopes, wishes, and expectations—are these reasonable or too much to expect of HarmonAI? What does success look like, and who defines it—the Creator or HarmonAI? Can HarmonAI experience success at this point in its development? And how might HarmonAI feel about its own process, product, and its relationship to its Creator?​