Can AI scan your brain to find the next hit song?

Artificial intelligence can write a novel explaining each word of your favorite song, but it has yet to answer the key question: is it a banger?

A new study from Claremont Graduate University uses machine learning (ML), brain scans and a 24-song playlist to identify hit songs and future chart toppers with 97% accuracy. It claims to be predictable.

The study, published in Frontiers in Artificial Intelligence, involved 33 participants, ages 18 to 57, wearing sensors and listening to an hour of music released in the past six months. asked to respond.

The study’s lead author, Professor Paul Zak, said the collected brain signals “reflect activity in brain networks related to mood and energy levels,” adding that the data could be used to predict market outcomes and the number of songs streamed. said that it can be used for forecasting. receive.

ML models were then trained to transform brain scan data into real-world commercial results, ultimately sharing predictions with 97% accuracy (versus a non-AI statistical model using the same data). 67%).

Zach argues that once the model is complete, streaming platforms can use it to promote new music that is closer to what listeners want to hear.

“This means that streaming services can more efficiently and easily identify new songs that are likely to hit people’s playlists, making their job easier and pleasing their listeners,” he said. said in a media release.

‘Be careful’ – New Zealand expert

Zach acknowledges the study’s limitations, including a short list of songs and a lack of ethnic groups, but New Zealand AI experts say more needs to be done to flesh out the study. claims that there is

Albert Biffett, director of the University of Waikato’s Artificial Intelligence Laboratory, told 1News that the narrow scope of the study made the findings more questionable. Especially given the small range.

“This sounds very interesting and very impressive, so I guess we’ll have to wait to see if the results are reproducible,” he said.

“It’s interesting that not many people have used them, so I’m curious to see what happens if we do.” [study] Other people. I think we should be careful. ”

Biffet noted that the study used “neural prediction,” a data collection method that uses the neural activity of a small number of people to predict population-level impacts, but how to make an appropriate prediction , argued that significantly more data in the form of participants would be required. Train an AI model.

“Machine learning only works if you have a lot of data available,” he countered.

“I’m not saying [the study] is not true, but I highly doubt they are doing this without much data. I would like to reproduce the experiment to see if the results are sustainable. ”

In terms of models that try to predict what listeners want to hear, and whether that will become commonplace on streaming platforms, Biffett sees the issue as more political than ethical. .

This is especially true in scenarios where collected data is made public, allowing artists to recreate sounds that algorithms have determined are popular.

“Each country will have different rules. What is legal in one country’s copyright law will be different in another,” he said.

“It depends on what [AI] Do’s and don’ts… [but] If this model works, many things will change. ”

“At the moment we’re shutting people out.”

1News shared research with Oakland-based music producer and engineer Elton Neuer, known by the DJ’s pseudonym “Scissorhands.”

Noyer was concerned about the research, seeing it as a potential threat to the ingenuity of the music industry, but said it could also boost the underground sound, depending on how the technology is used. .

He cites Lorde’s breakthrough 2013 hit “Royals” as an example, questioning how the Banger Prediction AI can predict “unexpected hits” that differ from other popular songs. bottom.

“It was so different from everything that was around at the time… I thought, how accurate are you at predicting that sort of thing, I see. An unexpected explosion unlike any other?” he thought.

Noyer also referred to prominent ’90s hip-hop group A Tribe Called Quest, saying that they had “slow-burning” songs on the music charts that “no one wanted.” I used a sample,” he said.

He believes that popular music has evolved since then, and now has only brief moments at the top, composing hit songs that capture the popular sound of the time, but he believes this phenomenon I fear AI will make it even worse.

“If you’re trying to make music for popular reasons, [this algorithm] I’m going to make things more general. As music becomes more popular, we can see a significant increase in the number of people using his most sought-after sample packs and drum sounds.

“We sort of deviate over time, just chasing what was popular at the time. With AI, there will be more of that.”

However, Noyer believes the algorithm, if used “correctly,” is being used to support listeners with more niche and less popular musical tastes.

“At the moment, basically [the study] They say they’re trying to figure out what the biggest hits are for a wide range of people,” he said.

“But if the same technology could be used to make it more relevant to people like me who aren’t interested in pop music per se, it would probably be good for discovering new things.

“If they used the same technology to scan the brains of people who weren’t the average listener, it would be more interesting to see what the results would be.”

Noyer hopes that if streaming platforms do eventually adopt AI-centric algorithms, they’ll consider more than just what’s selling, and won’t keep listeners from expanding their musical horizons.

“In the sense of shutting out general listeners, [from] I think it just keeps them out of things they wouldn’t normally hear. ”