Google has developed a new AI system called MusicLM that is capable of generating music in any genre given a text description. However, the company has no plans to release the system due to ethical concerns.
MusicLM was trained on a dataset of 280,000 hours of music to learn to generate coherent songs for descriptions of “significant complexity” such as “enchanting jazz song with a memorable saxophone solo and a solo singer” or “Berlin ’90s techno with a low bass and strong kick.” According to an academic paper detailing the system, MusicLM’s songs sound similar to what a human artist might compose, although not necessarily as inventive or musically cohesive.
MusicLM can also take several descriptions written in sequence and create a sort of melodic “story” or narrative that can be used for movie soundtracks. Additionally, the system can be instructed via a combination of picture and caption, or generate audio that’s “played” by a specific type of instrument in a certain genre. The researchers also note that the system can create music inspired by places, epochs, or requirements.
However, the system is not perfect, and some of the samples generated have a distorted quality. The researchers also found that about 1% of the music generated was directly replicated from the songs on which it trained, which is a high enough threshold to discourage them from releasing MusicLM in its current state.
The researchers also acknowledge the ethical challenges posed by a system like MusicLM, including the tendency to incorporate copyrighted material from training data into the generated songs. This has led to concerns about legal issues and potential copyright infringement. Critics have also questioned whether training AI models on copyrighted musical material constitutes fair use.
Overall, while the technology is impressive, there are many considerations and potential risks that need to be addressed before such a system is made available to the public.
