Music created by artificial intelligence is getting better and better, and musicians are worried.
Photo: Erwi / Unsplash
Your next favourite artist could be AI.
On today's episode of The Detail, AI music expert Ollie Bown, an associate professor at the University of New South Wales School of Art and Design, talks through how AI music platforms work, and the pros and cons of taking music creation out of human hands.
"When people say it's a revolution as big as electricity or the motorcar, they're not exaggerating," he says.
"In the last 10 or more years, there's been this relentless surge of progress [and] technical advance in the capability of AI systems. The most obvious evidence of that is in the language model stuff - Chat GPT and those AI language models - because they are so startlingly lifelike and humanlike in their response and in their ability to do basic logic.
"In music, we've got - quite surprisingly I think even to the experts - these systems that are able to produce original material at an audio level, which means that it's literally building, sample by sample, the audio waveform.
"It can simulate acoustic sounds, instrumental sounds, electronic sounds, voices and also the musical structure and even the long-term musical form that makes up a song - all of that's starting to be really successfully captured in AI."
The two sites that have garnered the most attention recently are Suno and Udio, both of which launched in the last year. These websites allow users to create an account and start generating music through a text description, in the space of a minute or two.
"What they both do very well is they integrate lyric generation and then the singing - the rendering of those lyrics into song - with the music generation, all integrated into one big black box, which is a phenomenal feat."
Bown talks through the advantages of using them. For instance, he says, AI could be good for sample generation.
"Say you're a hip-hop producer and you're looking for a hook to put some beats over and then rap over - you can have a great deal of fun with this, because you're already in that mindset of sampling, of lifting existing music and re-interpreting it and this just gives you a huge playground of songs and timbres and musical styles."
But there are also downfalls. At the forefront are concerns around intellectual property and copyright issues.
"Copyright law has just not been set up to account for this particular situation," Bown says.
"What the AI systems generate may very often sound very similar to something you've heard before, but not identical and it may be really blatantly in the style of a certain musician.
"The question is actually whether the AI companies have broken any law by gathering all of that data in the first place and feeding it into their model. It's hard to really pin down exactly what law they've broken."
Also in today's episode, RNZ senior studio operator Rangi Powick helps create some AI tracks - specifically about The Detail and its team members. Listen to the podcast to hear them.
Check out how to listen to and follow The Detail here.
You can also stay up-to-date by liking us on Facebook or following us on Twitter.
