SEOUL, September 14 (AJP) - Pitchfork, one of America's best-known music publications, does not hand out zeroes often.
But in August, American rapper Tyga's "$TARFACE" got one.
The album had already drawn attention after Tyga acknowledged using artificial intelligence as a production tool, including for 1980s-style synthesizers and guitar solos, while saying the lyrics and vocals remained his own.
Pitchfork did not reject the album simply because AI was involved. Its criticism was that the music sounded so generic and mechanically assembled that an AI generator might have produced something similar.
In Korea, the opposite happened.
"Maeumui Sup," or "Forest of the Heart," by independent creator Joseon HipHop entered Melon's Top 100 in early September and climbed as high as No. 89. The track pairs laid-back boom-bap drums with wistful, traditional-leaning melodies. The creator wrote the lyrics before using AI music generator Suno to help produce the composition and vocals.
The contrast points to a harder question.
If listeners enjoy a song made with AI — and may not even realize AI was involved — what exactly are they rejecting when they say they do not want AI music?
Bae Ji-hoon, a 28-year-old office worker, sees little reason to make the distinction.
Bae said he regularly listens to HanGroove, whose songs blend contemporary styles with sounds and lyrical elements reminiscent of traditional Korean music.
"When you first hear the phrase 'AI music,' you imagine that it is going to sound strange," Bae told AJP. "But when you actually listen, there are already a lot of songs that are really well made."
What matters to him is whether the song works.
"They hit my taste almost perfectly," he said.
Bae said early AI-generated songs often sounded noticeably unnatural, but the technology has improved quickly.
"Now there are times when I listen and wonder whether it really was made by AI at all," he said.
Listeners may not be very good at telling the difference.
Deezer and polling company Ipsos asked 9,000 people across eight countries to distinguish human-made music from fully AI-generated tracks. Participants heard three songs — two generated by AI and one made by humans.
Ninety-seven percent failed to identify all three correctly.
Yet 80 percent said fully AI-generated songs should be clearly labeled. More than half said they should not compete with human-made music on the same main charts, while 40 percent of streaming users said they would skip an AI-generated song without listening if they knew what it was.
The divide is emerging as AI music proliferates.
Deezer said fully AI-generated tracks reached about 90,000 uploads a day this summer and, at their June peak, accounted for more than half of all new music submitted to the service. Yet they generated only around 1 to 3 percent of total streams.
Listeners may struggle to hear AI, but they still want to know when it is there.
A study published in Psychology & Marketing this year found that nearly 88 percent of U.S. adults initially said they would prefer a human-made song to one generated by AI.
Researchers then played participants the same AI-generated song, telling one group it had been made by humans and another that it had been generated by AI.
Before listening, participants expected the supposedly human-made song to be considerably better. Afterward, the gap largely disappeared.
The song scored 6.92 out of 10 when labeled human-made and 6.56 when labeled AI-generated, a difference that was no longer statistically significant.
The label still mattered, however. Participants who believed a human had made the song were more willing to listen again, search for more music by the artist and share it with others.
"Music is something the listener should enjoy," Bae said. "If the listener is satisfied, isn't that enough?"
He sees little ethical problem with AI music as long as its use is disclosed.
For musicians, the attraction can be more practical.
"The biggest benefit of making music with AI is convenience," he told AJP.
He normally experiments with different sounds, instruments and plug-ins while composing. Suno can compress part of that process into minutes.
"You can put in lyrics or describe the mood you want and get a complete song unbelievably quickly," he said. "And you can keep generating it again and again."
Speed, however, does not necessarily produce a finished record.
Frankthehurman said AI-generated vocals can still sound awkward, particularly in pronunciation, and he would revise the output before using it in his own work.
"I might add more layers, or if it was a vocal recording, I would rather record it myself," he said.
For him, AI is closer to another production tool than a replacement for the musician.
"If more people become interested in composing or writing songs because of AI, I think that's a good thing," he said.
That also makes "AI music" increasingly difficult to define.
One musician may use AI for a single element while retaining control over the writing, arrangement and performance. Another may enter lyrics and a few instructions and receive an almost complete song.
Modern music already involves layers of human collaboration. Songwriters, topliners, arrangers, producers, vocal directors, engineers and performers may all contribute to a single release, sometimes across different countries and companies.
AI is now entering that chain, forcing copyright systems to decide how much human contribution is enough to constitute authorship.
South Korea confronted that question this summer.
The Korea Music Copyright Association introduced rules on Aug. 3 that would have allowed AI-assisted works to be registered when a human lyricist, composer or arranger had made a substantial and leading creative contribution.
It withdrew the change just 22 days later, saying broader discussion was needed over how human creativity should be recognized when AI is involved.
AJP Takeaways
- AI-assisted music is entering mainstream charts, but critical reception remains divided, as shown by Joseon HipHop's Melon entry and Pitchfork's 0.0 review of Tyga's "$TARFACE."
- Listeners often struggle to distinguish AI-generated music from human-made tracks, yet most still want clear disclosure when AI is involved.
- Research suggests AI can match human-made music in immediate listening appeal, but human authorship still boosts willingness to replay, share and follow an artist.
Copyright ⓒ Aju Press All rights reserved.