The Central Conservatory of Music in Beijing, not Shanghai, launched China’s first doctorate in musical artificial intelligence in 2019 — and it recruits computer scientists. The final exams everyone talks about are documented nowhere. What is documented is who gets in, and that says more.
On 1 March 2019 the Central Conservatory of Music in Beijing published what looked like an administrative footnote: China’s first doctoral programme in «musical artificial intelligence and music information technology». Each candidate would get two supervisors, one musical and one scientific. That September the first four doctoral students arrived. None came from another conservatory. They came from Tsinghua, from Beihang, from Nanjing University and from the Stevens Institute of Technology in New Jersey. They were engineers and computer scientists.
That is the right place to start, because it is the only truly solid fact in a story that mostly circulates in distorted form. The artificial intelligence at work in these programmes is not a plugin that writes tunes so the student doesn’t have to. It is machine learning systems trained on large corpora of scores and recordings, which extract statistical regularities — harmonic progressions, timbral profiles, rhythmic structures — and reuse them to generate or transform sonic material. The people inside these programmes spend most of their time on datasets, annotation, network architectures and evaluation. The output may be a piece of music; the daily craft looks far more like machine learning research than like a nineteenth-century composer at a piano.
The people involved: a department built in a year
President Yu Feng has said preparation began in 2018 and that the department was formalised in 2019, organised around three centres: music therapy, electronic music, and musical artificial intelligence. Li Xiaobing heads it. In May 2019 the discipline was added to those designated as priority fields by the Beijing municipal government; the dedicated department came into being that July. These are the timelines of an administration that has made a decision and wants to see it standing, not of slow academic maturation.
The numbers stay small, and they should be stated plainly: from four admitted in 2019 to eleven in 2024. The first doctorates were awarded in July 2022, and across two cohorts roughly ten students have graduated. This is not a factory turning out synthetic composers. It is a research group of European proportions, with an integrated master’s-to-doctorate track and a selection process explicitly reserved for candidates with computer science and AI backgrounds.
The most quoted — and most misread — detail is a statement by Li Xiaobing: admission to the doctorate in electronic music composition no longer requires the traditional harmony and orchestration examinations, but computer-based composition instead. Keep the proportions right. It concerns one specific track at one institution, and it is a spoken statement reported in the press, not a published regulation. This is not «the end of music theory in China». It is a sign that in that particular niche the entry filter has changed in kind.
Shanghai is doing something else, and it deserves its real name
The Shanghai Conservatory of Music is often described as the place where «AI-assisted composition» is taught, complete with dedicated final exams. The documentable picture is different and more modest. Its Music Engineering department has existed since 2003, with bachelor’s degrees in recording arts and in art and technology, and master’s tracks in electronic music design, music technology, recording and film music. The only doctoral direction tied to artificial intelligence is «musical AI». Its founding act — reported as 2020 — rests on a single press source; the official document is not retrievable.
Shanghai does, however, hold a national ministry project under the «new humanities» banner, dedicated to exploring and practising a musical artificial intelligence course. It is led by Yu Yang, who heads the department, served as musical director of the «AI Concert» at the 2019 World AI Conference in Shanghai, and mentored Microsoft’s Xiaoice system for the 2020 conference anthem. That is the concrete part: a funded programme of study on the subject, an identifiable person running it, a track record of public events.
In 2025 the conservatory launched a micro-specialisation called «Music, Science and Innovation». The details tell you a lot: thirty-five places, no tuition fee, applications from 1 to 5 September, interviews on the 8th, evening classes from the 15th. It awards a certificate, not a degree. Half the teaching staff comes from industry, with more than thirty corporate mentors, and the programme includes AI-assisted instruction. It is not a composition course: it is an interdisciplinary track spanning AI, virtual reality, smart manufacturing and entrepreneurship. Students attend in the evening, after regular classes, assembling a hybrid profile to take to the job market.
There is also a laboratory on AI and music therapy directed by Liu Hao, and a forum held on 22 October 2025 in the conservatory’s opera house, where AgiBot’s «Yuanzheng» robot, trained on the pipa repertoire, was presented. The same caveat applies: a robot learning a specific instrumental repertoire is an experiment in motor control and gesture modelling, not a machine that composes.
The gap sits exactly where everyone wants to look
What students actually submit for their final examinations — in Shanghai as in Beijing — appears in no public source. There are no exam syllabuses, no sample assignments, no lists of defended theses. It is the single most desirable piece of information and the only one missing. Anyone telling you that Chinese students hand in model-generated pieces at their finals, and that a panel grades algorithmic output, is filling that gap with imagination.
The same caution applies to two phrases the Beijing institution has circulated and the press has amplified: «China’s first AI symphony» and «the world’s largest music data centre». These are self-declarations by an interested party, with no independent verification. The surrounding infrastructure, by contrast, is verifiable — and it is the interesting part: an in-house AI composition system that produced the pieces «Welcome» and «A Thousand Li of Rivers and Mountains», a conducting robot named «Zhiyin» developed with Unitree, participation in a contemporary Chinese music festival at Carnegie Hall in October 2024, a joint laboratory with Huawei, and a supercomputing centre funded by the National Development and Reform Commission that came online in 2024. That is the project’s real physiognomy: hardware, industrial partners, compute.
And Europe? Hardly standing still
The claim that this has no equivalent outside China does not survive contact with the facts. The ATIAM master’s, run by Sorbonne Université with IRCAM and Télécom Paris, has for years trained people straddling acoustics, computer music and composition; IRCAM’s Cursus programme works precisely on composition and computer music. In London, Queen Mary University hosts the UKRI Centre for Doctoral Training in AI and Music, funded in 2019 — the same year as Beijing — offering a four-year doctorate with more than thirty supervisors, directed by Simon Dixon.
The difference is not whether the field exists, but how it is governed. In China the spread came through institutional channels and in a handful of years: Beijing and Shanghai from 2019, Sichuan in postgraduate courses from 2021. And one boundary clarifies the scale of the phenomenon: the 2025 ministerial catalogue of undergraduate degree programmes lists 845 entries, twenty-nine of them new, including «intelligent audiovisual engineering», «digital theatre» and «AI for education». A national undergraduate degree in AI composition is not among them.
Which brings us to the least intuitive fact in the whole affair. The change is not arriving through the front door of nationally approved degree curricula. It is coming in through the windows: research doctorates with about ten graduates in total, evening certificates with no legal standing, laboratories with corporate partners. Anyone who wants to understand where Chinese music education is heading should stop looking for the decree and start looking at who signs the mentors’ contracts.