Artificial intelligence (abbreviated as AI) will be the protagonist of the technological development of the next decades but more generally it will impact on different areas of society, work and organizations. All international analysts are giving great importance to research and applications of AI for development not only in the industrial field but also to support decisions, diagnostics and more generally choices. Since last year ISIA, the state university-level institute of Florence that offers three-year and specialist courses in product and communication design, has launched the “AI ISIA” project which was the first research laboratory in the field of AI applications. in design, communication and the arts.
With this aim, a Scientific Committee has been set up, which I have the honor to chair, which includes among its members researchers from important international companies, public and private research laboratories and universities. The aim is to explore, experiment and apply AI in the various disciplines of creativity and communication, develop degree and research theses, function as a reference point for the world of AI at an Italian and international level in these sectors.
At the end of the semester, a seminar was organized by ISIA and AI ISIA together with researchers, musicians, students and with the presence and participation of the musician Marco Castoldi in art Morgan, the research and some experiments were presented to the world for the first time.
The jazz master Antonio Gallucci he duet in real time with Obernaut01, an artificial intelligence designed and developed by the researcher and composer Francesco Bardozzo in art Oberlunar member of the Neurone Lab of the University of Salerno and member of the Scientific Committee of AI ISIA. The event, introduced by Prof. Francesco Fumelli, Director of ISIA Firenze has constituted an absolute primacy of an artificial intelligence that generates new compositions in real time and duets with musicians.
Antonio Gallucci underlined how improvisation had a very fast learning curve and how, despite having been a test of high difficulty, Obernaut01 learned quickly. Francesco Bardozzo has been working on the project for some time and until now he had experimented in public only a few fragments but which were sufficient to create an attention of the public and critics. The experiment is indicative of this new artistic relationship between human and artificial intelligence and was a first, but significant, step forward in the research of AI applications and how this project has an ambition aligned with the capacity of the team that is operating.
Bardozzo, aka Oberlunar, already known in the music world thanks to his recent experimental piano and AI album (Adhal, 2020) stated that “cThis experiment established the first true harmonic / melodic contact of a neural network with a human being in real time. In terms of both significance and musical depth, Antonio Gallucci’s improvisation with Obernaut01 was in some ways similar to Joyce’s epiphany. “
Francesco Bardozzo, who has always designed his musical-neuronal models inspired by Douglas Hofstadter’s research on strange rings, looking at some cardinal concepts on the self-referentiality, reflexivity and anti-parallelism of musical language, underlined how the preview of the The Jorgan File by Morgan, a multi-modal and as yet unpublished research on the science of modern composition was inspiring.
At the end of the performance, Gallucci explained how in improvising with Obernaut01 he had to radically change the way of thinking about the notes he produced, shifting the relations with the tonal center to different colors. He underlined how the result obtained goes beyond the jazz avant-garde, triggering alienating feelings. The common thought is that Obernaut01 did not produce serial or sterile music at all in terms of awareness and consciousness. The impression was that artificial intelligence has conceived its own way of producing music through a complex reasoning and, often, difficult to understand through the human filter. In technical terms, Francesco Bardozzo defines the musical reasoning developed in AI as scientifically defined in its so-called latent space, similar to a world of ideas.
Francesco Bardozzo also emphasizes the technique used “I trained the neural network with pieces by Scott Joplin, Louis Armstrong and Michel Petrucciani but I could have trained it on Johann Sebastian Bach or on Somewhere else by Morgan. The adaptive and generalizing power of this network makes it ready for any musical genre “. Beyond the success established by this first contact, Morgan and Oberlunar questioned the possibility that Obernaut01 could generate a sort of emotional transference in the listener. Obelunar says that “in various comparisons with Morgan, we have developed a procedure to evaluate the degree of musical intelligence generated by the machine. This type of test exists in a more general form but has never been defined with regard to musical semantic language. “
In the past, other experiments were carried out but the idea behind the Obernaut model is to create a self-supervised or unsupervised system. In addition to the training dataset, which is human, it is the network itself that composes music, creating its own harmonic-melodic rules. In previous experiments it was the neural network that had to identify the musician’s rules and follow his suggestions. His goodness was valued on how close he was to the idea of the musician himself. In this experiment, however, the paradigm is reversed: it is man who has to adapt and change his musical rules to be able to duet with the car.
Another element to take into consideration is that most of the previous experiments, although they have elements of AI, maintain the typical structure of random music. In this case Obernauts 07 is a system of several AI that cooperate in unison to define not only the sequence of the notes, but also their intensity, speed and rhythm, as a musician would do in the compositional and expressive act, avoiding clearly the randomness.
The event was actively attended by the musician and singer Morgan, who underlined how experimentation is not only an opportunity to reconsider musical dynamics but more generally to study, compose and break down the different methods for imagining, writing and producing music. The teacher Igor Merlini, composer and RAI consultant for music and also a member of the Scientific Committee of AI ISIA, upon first listening to the products of Obernaut01, argues that in the near future the generative composition based on neural networks will be able to change the musical creative process, subverting many paradigms of modern composition. The teacher Nicola Elias Rigato, composer and close collaborator of Oberlunar for the Obernaut01 project, explained that the composer’s role will be to train and instruct the neural network, thus creating everything that is upstream and not downstream of the composition itself.