The organizing committee of the 5th China in My Eyes (CiME) Short Video Contes

The song lasts three minutes and three seconds, and every second of it was argued over. The opening line sets up the whole idea: “I turn my head to view, two million stories coming through; from people everywhere, we film the moments that we share.” The two million stories are the foreigners living, working and creating in China — each of us a camera, each of us a narrator. The chorus turns the camera around: “See me, as one! Singing together, everyone!” The title works in both directions at once. China is what I see; but as a foreigner who has made a life here, I am also something China sees. Being seen, and seeing truly, are the same act performed from two sides.

“See me, as one! Singing together, everyone!”

Musically I wanted something a child could sing by the second listen and an adult would not tire of by the tenth, so the melody stays within an octave, moves mostly stepwise, and sits over a familiar four-chord loop at a brisk 125 beats per minute. The verses alternate English and Chinese freely — “China in my eyes, I see a bright light; singing our song, this is where we belong” — because that is how bilingual people actually think, and I saw no reason the song should pretend otherwise. There is a sixteen-bar instrumental in the middle, with crowd voices chanting the hook underneath, and I reserved the solo for my own guitar. Whatever else on the recording came from machines, those bars are mine.

That distinction mattered more than it might seem. The winning agreement grants the organisers the right to broadcast, re-edit and promote the work, which is entirely fair for a theme song — but it makes a nonsense of any entry built on a tool whose free tier forbids commercial use. So I did not press one button on a music generator and hope. Instead I composed with my own AI systems, developed and refined over three or four years, with lyrics drafted and redrafted in collaboration with several large language models — Kimi, DeepSeek, Gemini 4 Argon and Claude Opus among them — music generated through Google’s Lyria, vocals produced in ACE Studio, and the whole production assembled and mixed in Apple Logic Pro, with agentic coding environments such as Cursor, Trae and Antigravity doing the unglamorous technical work in between. If that sounds like a long list, it is; but then, so is the list of people behind any orchestra.

For the past two and a half years, as I built the systems behind this song, I have been fortunate to have my former tutor, Professor Matthew James Sansom, as a sounding board. Now Professor of Music and head of the music department at the Beijing Normal University–Hong Kong Baptist University United International College in Zhuhai, Matthew earned his doctorate at Sheffield studying free improvisation and musical meaning, spent two decades teaching at Newcastle and Surrey, and held a pro-vice-chancellor post at Sunway University in Kuala Lumpur before coming to China. He is a working artist in his own right — saxophonist, electronic musician and installation artist — whose work has been heard at the Victoria & Albert Museum, the ICA in London and the National Theatre of Prague, and at UIC he now leads the school’s AI-literacy and music-technology initiatives. His belief, in his own words, is that education should make people “more independent thinkers, more critical, more creative” — which is a fair summary of what he has asked of me throughout this work. A model can suggest a hundred choruses; it cannot tell you which one is true.

I will admit to a professional anxiety about the whole enterprise. I teach music; I also sit on our school’s Theory of Knowledge group, where we spend staff meetings asking how we know what we know, and whether a machine can know anything at all. Writing a song with AI was, in a sense, setting myself a TOK essay in public. What I concluded is this: the tools handled tasks — harmony suggestions, vocal synthesis, mixing assistance, code — while the judgement, the structure, the lyrics, the bilingual phrasing and the final responsibility remained stubbornly human. The contest judges, whatever their criteria, heard a song in two languages with a clear idea at its centre. The idea was the part no model supplied.

The ceremony itself was a fine reminder of what the song was about. Around me sat filmmakers from dozens of countries, all of us strangers who had nonetheless produced a small library of China as we experience it — its food, its neighbourhoods, its high-speed mornings and its quiet courtyards. When “See Me” played for the first time in that hall, it did not feel like my song. It felt like the room’s song, which is precisely what a theme song is for.

If there is a lesson I can bring back to my classroom, it is that our students should not fear these tools, nor worship them. The musicians who will thrive are the ones who can state clearly what they mean, and then delegate everything else. That, and the older lesson underneath it: two million stories are coming through every day in this country. All we have to do is turn our heads to view — and then, in our own voice, sing.

With thanks to Dr Max Caruso (FIoL), Co-Principal of Wuxi United International School, whose encouragement gave me the opportunity to write this song.

At UIC he launched the school’s AI Literacy seminar series (“What is AI Literacy and Why it Matters?”, November 2024) and leads its Strategic School Sound Initiative on the future of AI and music.