History of Algorithmic Music Generation: From Mozart to Modern AI
Abstract
Algorithmic Music Generation is an old topic. From very basic systems of controlled randomness to advanced data-driven systems. This paper summarizes the general history of algorithmic music generation, as well as three key eras: the random combination era, the rule-based algorithm generation era and the deep learning era. Mozart's Musikalisches Würfelspiel (meaning musical dice game) is the first stage. This is a simple example of using predefined music materials for music generation, and random selection. The second stage was ushered in with the advent of electronic computers and algorithms such as Markov-chain and stochastic modeling. In this period the musical elements of pitch, rhythm, harmony and orchestration became formalized as parameters for computation. Even so, they were still limited in their creative abilities by rules set forth explicitly by humans. Third stage is based on deep learning. In this time period, algorithmic music generation shifts to data-driven modeling. For systems that learn directly from large amounts of data, musical patterns are learned using transformer architectures, diffusion models, and cross-modal alignment techniques, and the generation of symbolic music or high-quality audio is based on musical or natural-language inputs. In this review the mathematical and computational foundations are explored for each era, along with the shift in the role of humans to computational systems. Additional difficulties, such as long-range structural coherence, controllability, and the assessment of originality are also discussed.
References
- Hedges, S. A. (1978). Dice music in the eighteenth century.Music & Letters, 59(2), 180–187.
- Mathews, M. V., & Miller, J. E. (1981).The technology of computer music. MIT Press.
- Hiller, L., & Isaacson, L. (1993).Musical composition with a high-speed digital computer.
- Hiller, L. A., & Isaacson, L. M. (2012).Experimental music: Composition with an electronic computer. Greenwood Publishing Group Inc.
- Xenakis, I. (1992).Formalized music: Thought and mathematics in composition. Pendragon Press.
- Eck, D., & Schmidhuber, J. (2002). A first look at music composition using LSTM recurrent neural networks. InIstituto Dalle Molle Di Studi Sull Intelligenza Artificiale, 103(4), 48–56.
- Huang, C.-Z. A., et al. (2018). Music transformer.arXiv Preprint, arXiv:1809.04281.
- Sexton, T. (2023). MuseNet.Music Reference Services Quarterly, 26(3–4), 151–153.
- Borsos, Z., et al. (2023). AudioLM: A language modeling approach to audio generation.IEEE/ACM Transactions on Audio, Speech and Language Processing, 31, 2523–2533.
- Agostinelli, A., et al. (2023). MusicLM: Generating music from text.arXiv Preprint, arXiv:2301.11325.