AI in music production: the impact of innovative tools like Google’s MusicLM are limited at the moment, but the potential is there

The accelerated development of artificial intelligence (AI) tools might leave a significant imprint on the music industry in the near future. Pioneering AI-powered systems such as Google’s MusicLM are offering new opportunities for musicians and producers to innovate and create, streamlining workflows and extracting patterns from large musical libraries. Legal and ethical concerns surround AI music generators like MusicLM, which potentially violate music copyright by creating coherent audio from ingested works during training. Critics have questioned whether training AI models on copyrighted material constitutes fair use, and similar concerns have been raised regarding training data used in image-, code-, and text-generating AI systems, often obtained from the web without the creators’ knowledge.

Researchers at Google recognize the ethical challenges posed by systems like MusicLM, including the inclination to incorporate copyrighted material from training data into the generated songs. As a result, the company has no immediate plans to release the system to the public. And despite its capabilities, MusicLM exhibits some technical shortcomings too, such as distorted audio quality and suboptimal vocal generation.

But what is Google’s MusicLM actually? Well, it’s an advanced AI system, designed to generate musical compositions from text prompts. So far MusicLM has been able producing minutes-long musical pieces from various input types, including descriptions, genre, vibe, or specific instruments. This AI system is built on the Transformer architecture, utilizing self-attention mechanisms and deep learning techniques to create high-quality music.

The benefits of employing MusicLM in music production include its superior performance compared to other music-generating AI systems, both in terms of quality and adherence to input captions. Additionally, its ability to replicate melodies and the release of Google’s dataset, comprising approximately 5,500 music-text pairs, provide valuable resources for researchers and developers looking to train and evaluate other musical AIs.

This is the link for the MusicLM project demonstration.

MusicLM is not the first generative AI system for music composition. Other attempts include Riffusion, Dance Diffusion, and OpenAI’s Jukebox. However, due to technical limitations and restricted training data, these systems have not been able to produce particularly complex or high-fidelity compositions.