André Holzapfel

André Holzapfel, Petra Jääskeläinen, Anna-Kaisa Kaila

The recent developments of artificial intelligence increase its capability for the creation of arts in both largely autonomous and collaborative contexts. In both contexts, Ai aims to imitate, combine, and extend existing artistic styles, and can transform creative practices. In our ongoing research, we investigate such Creative-Ai from sustainability and ethical perspectives. The two main focus areas are understanding the environmental sustainability aspects (material, practices) in the context of artistic processes that involve Creative-Ai, and ethical issues related to who gets to be involved in the creation process (power, authorship, ownership). This paper provides an outline of our ongoing research in these two directions. We will present our interdisciplinary approach, which combines interviews, workshops, online ethnography, and energy measurements, to address our research questions: How is Creative-Ai currently used by artist communities, and which future applications do artists imagine? When Ai is applied to creating art, how might it impact the economy and environment? And, how can answers to these questions guide requirements for intellectual property regimes for Creative-Ai?

Petra Jääskeläinen, Daniel Pargman, André Holzapfel

An increasing number of artists use Ai in their creative practices (Creative-Ai) and their works have by now become visible at prominent art venues. The research community has, on the other hand, recognized that there are sustainability concerns of using Ai technologies related to, for instance, energy consumption and the increasing size and complexity of models. These two conflicting trajectories constitute the starting point of our research. Here, we discuss insights from our currently on-going fieldwork research and outline considerations for drawing various limitations in sustainability assessment studies of Ai art. We provide ground for further, more specific sustainability assessments in the domain, as well as knowledge on the state of sustainability assessments in this domain.

Guillermo Valle-Pérez, Gustav Eje Henter, Jonas Beskow et al.

Dance requires skillful composition of complex movements that follow rhythmic, tonal and timbral features of music. Formally, generating dance conditioned on a piece of music can be expressed as a problem of modelling a high-dimensional continuous motion signal, conditioned on an audio signal. In this work we make two contributions to tackle this problem. First, we present a novel probabilistic autoregressive architecture that models the distribution over future poses with a normalizing flow conditioned on previous poses as well as music context, using a multimodal transformer encoder. Second, we introduce the currently largest 3D dance-motion dataset, obtained with a variety of motion-capture technologies, and including both professional and casual dancers. Using this dataset, we compare our new model against two baselines, via objective metrics and a user study, and show that both the ability to model a probability distribution, as well as being able to attend over a large motion and music context are necessary to produce interesting, diverse, and realistic dance that matches the music.

Georgi Dzhambazov, Andre Holzapfel, Ajay Srinivasamurthy et al.

The goal of this study is the automatic detection of onsets of the singing voice in polyphonic audio recordings. Starting with a hypothesis that the knowledge of the current position in a metrical cycle (i.e. metrical accent) can improve the accuracy of vocal note onset detection, we propose a novel probabilistic model to jointly track beats and vocal note onsets. The proposed model extends a state of the art model for beat and meter tracking, in which a-priori probability of a note at a specific metrical accent interacts with the probability of observing a vocal note onset. We carry out an evaluation on a varied collection of multi-instrument datasets from two music traditions (English popular music and Turkish makam) with different types of metrical cycles and singing styles. Results confirm that the proposed model reasonably improves vocal note onset detection accuracy compared to a baseline model that does not take metrical position into account.