In the early 1980s the composer David Cope started Experiments in Musical Intelligence, known as EMI or “Emmy,” after composer’s block on an opera commission led him to program a system that could write music for him. EMI analyzed a database of a composer’s works and generated new compositions in the same style without exactly reproducing any of the originals.
EMI belongs to the symbolic, rule-and-pattern tradition rather than the neural one. It worked by breaking existing scores into recurring fragments and characteristic gestures, then reassembling them under stylistic constraints into novel pieces. Cope produced volumes of EMI compositions credited to a virtual composer, demonstrating that a program could capture enough of a style - Bach chorales, Chopin nocturnes, Mozart sonatas - to fool trained listeners.
The cognitive scientist Douglas Hofstadter used EMI in public lectures as a kind of musical Turing test, asking audiences to distinguish a genuine work from an EMI imitation; listeners often could not. The result anticipated the central tension of today’s AI music tools: if a machine can convincingly imitate a style, what remains distinctly human about composing, and who should be credited.
Why business readers should care: EMI is a decades-old proof that generative systems can master the surface of a creative style, framing the authorship and licensing questions that AI music products face now.