Voices from AI in Experimental Improvisation
Tomomi Adachi (JP), Andreas Dzialocha (DE), Marcello Lussana (IT)
Through machine learning, computers can recognize patterns in a variety of sound documents. But can they also learn to improvise musically? The software “Tomomibot” tries it out by interacting in real time with the sound artist Tomomi Adachi, using deep learning techniques. The Artificial Intelligence (AI) "learns" the artist’s individual style via voice recordings and directly confronts him with the newly generated material. Their joint performance shows how interactive technology and AI can influence a (vocal) style. However, this dialogue also makes clear that the artist will always be more creative and unpredictable than his mechanical counterpart.
