"Every day, thousands of songs are released. This constant influx of choices makes it difficult for streaming services and radio stations to determine which songs to add to their playlists. To find songs that resonate with the public, these services rely on human listeners and artificial intelligence. However, this approach only has an accuracy rate of 50%, making it unreliable in predicting whether a song will become popular.
Now, researchers in the United States have applied comprehensive machine learning techniques to brain responses, enabling them to accurately predict 97% of popular songs.
"By applying machine learning to neural physiological data, we can almost perfectly identify popular songs," said Professor Paul Zach of Claremont Research University, a senior author of the study published in the journal "Frontiers in Artificial Intelligence." "This research demonstrates that the neural activity of 33 individuals can predict whether millions of people will listen to a new song, which is incredibly surprising. Such high accuracy has never been demonstrated before."
Using neural system data for machine learning
Participants in the study used pre-existing sensor devices to listen to a set of 24 songs and were asked to provide personal preferences and some demographic data. During the experiment, scientists measured participants' neural physiological responses to the songs. "The brain signals we collected reflected the activity of brain networks related to emotions and energy levels," said Zach. This allowed researchers to predict market outcomes, including the number of streams, based on a small amount of data.
This approach is called "neural prediction." It predicts the effects on the overall population by capturing the neural activity of a few individuals, without measuring the brain activity of hundreds of people.
After collecting the data, researchers used different statistical methods to evaluate the predictive accuracy of neural physiological variables. This allowed for direct comparisons between models. To improve prediction accuracy, they trained a machine learning model and tested different algorithms to achieve the highest prediction results.
They found that linear statistical models identified popular songs with a success rate of 69%. When they applied machine learning to the collected data, the proportion of correctly identified popular songs jumped to 97%. They also applied machine learning to the neural responses of the songs' first minute. In this case, popular songs were correctly identified with a success rate of 82%.
"This means that streaming services can more efficiently identify new songs that may become popular in people's playlists, making their work easier and making listeners happier," explained Zach.
Replicating the method
"If in the future, portable neuroscientific technologies like the ones we used in this study become widespread, appropriate entertainment content can be sent to audiences based on their neural physiology. They may only be given two or three choices instead of hundreds or thousands, making it easier and faster for them to choose music they like," said Zach.
Despite achieving near-perfect prediction results, the research team pointed out some limitations. For example, they used a relatively small number of songs in their analysis. Additionally, the demographic characteristics of the study participants had moderate diversity but did not include members of certain racial and age groups.
Nevertheless, the researchers anticipate that their method could be used beyond identifying popular songs, partly due to its ease of implementation. Zak concluded, "Our main contribution is the methodology. This approach is likely to be applicable to predicting the popularity of many other types of entertainment, including movies and TV shows."
Source: https://techxplore.com/news/2023-06-machine-songs-accuracy.html"
Alphabet Q1 paid subscribers reached 350M, fueled by YouTube & Google One. YouTube ad revenue slightly missed at $9.88B amid Premium growth. Total Q1 revenue hit $109.9B, beating forecasts.
Anthropic releases Claude connectors for Adobe, Blender, and Ableton, enabling AI workflow automation and app control. It also sponsors the Blender Development Foundation.
Apple introduces an App Store model splitting 12-month plans into monthly installments. On iOS 26.4+, it lowers upfront costs without reducing total price, adding billing tracking.
Google launches Aadhaar digital IDs in Wallet for India, enabling secure QR/NFC verification. A phased partner rollout targets key sectors, while passport-based credentials expand globally.
YouTube tests "Ask YouTube," an AI tool blending text and video for complex queries. Initially for US Premium users, it supports interactive follow-ups, expanding Google's AI search.
Google grants US DoD classified AI access, joining OpenAI and xAI in defense deals amid Anthropic's dispute. Contractual safeguards against surveillance and lethal weapons remain unenforceable.
Mistral AI launches Workflows, a Temporal-powered orchestration platform for secure, Python-driven enterprise AI automation featuring human-in-the-loop approvals.