Am I (Deep) Blue? Music-Making AI and Emotional Awareness

Abstract

Artificial Intelligence has shown paradigmatic success in defeating world champions in strategy games. However, the same programming tactics are not a reasonable approach to creative and ostensibly emotional artistic endeavors such as music composition. Here we review key examples of current creative music generating AIs, noting both their progress and limitations. We propose that these limitations are rooted in current AIs lack of thoroughly embodied, interoceptive processes associated with the emotional component of music perception and production. We examine some current music-generating machines that appear to be minimally addressing this issue by appealing to something akin to interoceptive processes. To conclude, we argue that a successful music-making AI requires both the generative capacities at which current AIs are constantly progressing, and thoroughly embodied, interoceptive processes which more closely resemble the processes underlying human emotions.

Cognitive modeling, ecological psychology, and musical improvisation

“…For example, LIDA’s focus on AAs means that it does not reduce improvisation to a problem-solving task, since the explorations, sensations, and alterations that are part of being an AA covers a wide swath of why we improvise. The central focus on affect, emotion, and consciousness also makes LIDA an important architecture to use and further develop; indeed, while affect and emotion are an essential part of all human cognition, they play an extremely strong role in music ( Schiavio et al, 2017 ; Van der Schyff and Schiavio, 2017 ; Novelli and Proksch, 2022 ).…”

“…Similarly, work in artificial intelligence has been steadily advancing. Examples of notable AI programs here include Voyager ( Lewis, 2000 ), Magenta, and BachBot ( Novelli and Proksch, 2022 ).…”

Harmonizing minds and machines: survey on transformative power of machine learning in music

“…The authors (Tang et al, 2022) focused on applying deep learning (DL) and ML techniques in music education. The authors (Novelli and Proksch, 2022) explored the limitations of current music-generating AIs regarding emotional awareness and proposed potential solutions to address this issue, discussing the potential impact of emotionally-aware music-generating AIs on the music industry.…”