CCMSAI 2026

Artificial intelligence (AI) in music has a long history extending at least to the early experiments of Lejaren Hiller and Leonard Isaacson in the 1950s. In the 2020s the combination of powerful GPUs, large datasets, and pre-trained transformer models has produced a dramatic expansion of AI-based musical practice. This paper explores promising trajectories for AI-based musical co-creativity considering two conflicting realities. First, the ethical and technical challenges of musical AI are not entirely novel. They mirror issues embedded in early digital and pre-computational algorithmic systems. Second, musical AI research and development has undergone a shift in its center of gravity to the private sector as seen in companies like Suno, and Google’s Magenta project, due in part to greater corporate computational resources and gigantic datasets that may not always be fully compliant with copyright law. Given the disparity in computational and data resource levels between academia and industry, academic researchers are exploring alternative trajectories, such as establishing ethical guidelines, developing ethically sourced datasets, and creating music from artist-curated datasets. However, a foundational question persists in all these activities: is AI-generated music structurally distinct from music produced without these technologies? This paper nudges that question forward through a combination of historical reflection, musical and technological observations, reports on experiments with AI in music, and speculations about possible futures for AI and music.