MusicLM : Google’s Quill Bot for Music

What is Google’s MusicLM fading away & Why is the buzz for MusicLM fading away?

With the release of ChatGPT the heat and race to come up with innovative AI offerings has never been so exciting.

What is MusicLM?

A model generating high-fidelity music from text descriptions such as “a calming violin melody backed by a distorted guitar riff”. MusicLM casts the process of conditional music generation as a hierarchical sequence-to-sequence modeling task, and it generates music at 24 kHz that remains consistent over several minutes

MusicLM by Google

In simple terms, musicLM is able to generate music by giving simple text or references like that of electric guitar/flute or conditions like a combination of text and melodies.

Here are some music generated by MusicLM :

Difference from ChatGPT and Brad?

This is not the first time someone has tried to generate AI music, Riffusion has also been released in December 2022 but Ruffusion uses images to generate music rather than references, text or conditions that are used by MusicLM.

Music is a very complex art to master and is very tough to model. ChatGPT and Brad are in different fields i.e solving the everyday issues of random searching and giving precise answers to questions. MusicLM is dedicated to generating music or sounds or melodies that are very specific and very different from search result-focused ChatGPT and Brad.

Challenges and Limitations

• Replicating an Art: Music is an art, that is different from the search-focused ChatGPT and Brad, which is MusicLM trying to master. Music’s taste, history, melodies, flow, length, instruments and purpose change from country to country and even from county to county in culturally diverse countries like India, Thailand, South Africa or Brazil.
• Lack of training models: As music is created in sound studios that are scattered around the world, the data and art of creating are dispersed around the world.
• Human acceptance of music: Scientists and biologists are still understanding why some pitches, sounds and frequencies are liked by the human brain and some are rejected. Even the most renowned music composers are not aware of which songs will strike the masses.
• Music Memory: Human brain is complex and we relate music to memories & humans’ liking of music is difficult to decrypt. As memory for each individual is stored in brains and not present in data on the Internet, adding emotional memory to the AI model is going to be very difficult and time consuming.

Conclusion

It is difficult to predict where generative AI for music will take us in the future. One thing is certain: even bigger models using even bigger datasets and even sharper algorithms will replace and improve MusicLM. Many of MusicLM’s flaws will undoubtedly be resolved by these models.

The music industry will undoubtedly be significantly disrupted by innovations like this, most likely sooner rather than later. But I think it would be a mistake if we devoted all of our efforts to black-box models.

For the most part, the world does not require machines to produce music from beginning to end. Humans are here for that. What matters most is that we use artificial intelligence (AI) technologies to add more creative value.