AI and music

Artificial intelligence (AI) is now giving us the next generation of electronic music, this time by using machine learning to understand sounds, patterns and styles of music and lyrics, and then generating new versions.

This was the approach used to create ‘The official Tokyo 2020 beat’, which saw an Intel AI use thousands of pieces of music reflecting themes of sports, Japanese culture, daily life and nature, to compose hundreds of options, before the final version was chosen by the Japanese public.

AI music is becoming a new industry, with start-ups inventing faster and easier music composition software, such as JukeDeck, a UK-based AI music start-up recently acquired by TikTok, which automatically interprets video and sets music to it.

Many recording artists are experimenting with AI for lyric generation, following the example David Bowie set in the 1990s with the song Hallo Spaceboy.

Others are using AI to make new neural synthesiser sounds, such as Grimes on the track So Heavy I Fell Through the Earth – Art Mix. And then there are those using AI to help create entire albums, like Taryn Southern with I AM AI.

With AI advances showing no signs of slowing down, it won’t be long before a computer can be used to make new versions of every musical genre that are indistinguishable from human-composed pieces.

AI can even innovate, creating new concepts and exploring new sounds that have never existed before. But like our languages, music is all about communication. Today AI has all the skills, but nothing to say. Only when we use it to help us express ourselves can its true value be felt.

Some have even used AI to create ‘new’ music from the likes of Amy Winehouse, Mozart and Nirvana, feeding their back catalogue into a neural network. Music AIs use neural networks that are really large sets of bits of computers that try and mimic how the brain works. And you can basically throw lots of music at this neural network and it learns patterns – just like how the human brain does by repeatedly being shown things.

What’s tricky about today’s neural networks is they’re getting bigger and bigger. And they’re becoming harder and harder for humans to understand what they’re actually doing.

We’re getting to a point now where we have these essentially black boxes that we put music into and nice new music comes out. But we don’t really understand the details of what it’s doing.

These neural networks also consume a lot of energy. If you’re trying to train AI to analyse the last 20 years of pop music, for instance, you’re chucking all that data in there and then using a lot of electricity to do the analysis and to generate a new song. At some point, we’re going to have to question whether the environmental impact is worth this new music. A computer may be able to make hundreds of tracks easily, but there is still likely still a human selecting which ones they think are nice or enjoyable.

Science Focus