Does Art Need to Be Hard? Rosalía, AI, and the value of difficulty - diggitmagazine.com

Does Art Need to Be Hard? Rosalía, AI, and the value of difficulty

AI has become more and more prominent in the music industry. This article explores to what extent the perceived difficulty of the artistic process influences how listeners value music created with versus without artificial intelligence. To answer this question, it examines Rosalía’s studio album Lux in contrast to Haven’s I Run, which was taken down from streaming platforms amid allegations of copyright infringement and speculations about the vocals having been generated by AI.

We have arrived at an age in which AI art is present in our everyday lives. In the last few months of 2025, at least six AI or AI-assisted artists have debuted on various Billboard rankings. These figures could be higher since it is becoming increasingly difficult to tell who or what is powered by AI (Zellner, 2025).

In October of 2025, the song I Run by British artist Haven went viral on TikTok. Shortly after, it was taken down from streaming platforms amid allegations of copyright infringement. The song credited no vocalist, leading to speculation that the vocals had been generated with AI. Haven later stated that he wrote and produced the track himself but used AI-assisted vocal processing to transform his own voice. Following overwhelming criticism, the song was re-released with new vocals performed by another artist (Savage, 2025).

In contrast, Spanish artist Rosalía has explicitly emphasized that her new album was made without any artificial intelligence: “It’s all human – very much human” (Coscarelli & Caramanica, 2025). In early November, she released her fourth studio album titled Lux, on which she is accompanied by the London Symphony Orchestra (Petridis, 2025). On the album, she sings in thirteen different languages, including both of her native languages, Spanish and Catalan, as well as English, Latin, Arabic, and German (Coscarelli & Caramanica, 2025). The album mixes pop music with flamenco and neoclassical arrangements, and the thematic framework is built around the lives of various female saints (Stamp, 2025).

Rosalía has said that she demands a great deal from her listeners with her new album. She frames Lux as a reaction to the loss of focus encouraged by social media (Petridis, 2025). Literary critic Bhandari (2025) argues that pop music does not always demand much from the listener or from the artist. Pop, he says, is often designed for virality, for fragments, for predictable nostalgia that never sounds old. It is essentially algorithm pop: built to catch the scroll and fuel the loop. In Haven’s case, the use of AI allows him to achieve a desired vocal style without collaborating with another artist. In contrast, Rosalía devoted herself fully to the production process. She spent two years creating the album and learning to write and sing convincingly in other languages (Coscarelli & Caramanica, 2025). The album also uses zero loops, includes breath intakes, and preserves the natural sound of strings (Bhandari, 2025). Rosalía describes this in an interview with radio DJ Zane Lowe (Apple Music, 2025, 15:36): “I wanted that it was acoustic instruments […] So at the end of the day, you have the sense, the feeling, that there is a human in there. It’s made by a human.”

These two very different musical projects have both found commercial success, raising questions about music and the role of artificial intelligence. When comparing them, the concept of difficulty becomes central. Both the difficulty of producing a piece of art and the difficulty of fully engaging with it. So, does art need to be difficult? And what do we lose or gain when we allow AI to assist in, or fully create, our art?

This leads to the main question of this article: to what extent does the perceived difficulty of the artistic process influence how listeners value music created with versus without artificial intelligence?

This article will explore this question by further examining Rosalía’s Lux in contrast to Haven’s I Run It uses Kraaijeveld’s (2024) understanding of the meaning of art and Salvaggio’s (2025) concept of interpassivity to explore the relationship between art and difficulty.

Back to topDefining art

To distinguish AI-generated art from human created art, philosopher and ethicist Steven Kraaijeveld (2024, p. 217) discusses three criteria to define a work of art: meaning, embodiment, and interpretation. He argues that AI-generated works can satisfy the criteria of embodiment, since machines can paint on canvases or generate novels that are then printed and bound. The third criteria could also be fulfilled, some people may interpret AI-generated pieces as legitimate works of art.

However, Kraaijeveld (2024) states that it is necessary for art to have meaning to be considered art. He uses an example to make his point (p. 217), by saying that a splash of paint needs to be intentionally placed instead of accidentally spilled to create meaning. The splash might look beautiful when accidentally spilled, but without intention there is no meaning. He brings this back to AI-generated art by saying that with the absence of meaning, the work must be restricted to meaning-as-interpretation. This means that meaning is projected onto the work by the viewer rather than inherent to the act of creation.

However, when it comes to AI-art generation, prompts created by human users are intentional. This raises the possibility that AI-generated art contains intention-based meaning through the choices of the prompter. Yet Kraaijeveld (2024, p. 218) distinguishes between the intention behind commissioning a work and the intention behind creating one. He writes: “If I ask a painter to paint my portrait, the portrait’s meaning is connected to what the artist chooses to create—e.g., how she expresses my existence and personality through paint on a canvas. That I have commissioned the portrait is important for its existence, but that is not something that we usually count in determining the meaning of art.” In other words, meaning is shaped by the artist’s deliberate choices, not merely by the commissioner’s request. This is a central difference between AI-generated and human-created art.

Back to topThe Value of the Artistic Process

As seen in Haven’s case, the artist did not rely on AI to generate an entire song. Rather, he used AI-assisted processing to achieve a particular vocal sound. So why did he receive such intense criticism? This brings us to the value we place on the artistic process itself.

Participation is a form of effort. Allowing AI to create or partially create art so that a person does not have to engage in that effort is an example of interpassivity: the outsourcing of activity, creativity, or even pleasure to machines (Salvaggio, 2025). This contrasts with the idea that the artistic process has been democratized. Now more than ever before, people have access to artistic tools. However, it seems that people are overwhelmed with the demand to produce, respond, engage, and perform. Salvaggio (2025) describes this as a deep cultural exhaustion surrounding participation. He writes that the demand for participation means that when we create something, we have to contort ourselves to algorithmic curation of what gets seen. Even though this does not accurately represent success, likes and views are how we measure popularity. As a result, unique or ambitious creations are often ignored, not because they lack value, but because they do not fit the easily consumable patterns rewarded online.

Nguyen (2025) echoes this concern, writing that the online ecosystem is optimized for viral, low-effort content. She argues that the value of genuine artistry has become more urgent, because it can restore the capacities that the attention economy erodes, namely patience, contemplation, and sustained focus. Rosalía’s Lux and Haven’s I Run highlight this tension.

Steiner (1978) reminds us that difficulty is inherent to both creating and experiencing art. Artistic choices, and thus participation, are embedded with intention and skill. As mentioned before, this is what Kraaijeveld (2024) says differentiates meaning in AI-generated and human-created art. Rosalía uses what Steiner (1978, p. 270) calls tactical difficulty. With her new album, Rosalía tries to slow down the listener and provoke thoughts. In an interview she says: “The more we are in the era of dopamine, the more I want the opposite” (Coscarelli & Caramanica, 2025). She insists on meaningful difficulty, intentional skill, and immersive artistic labor.

Haven, in contrast, leans into the conveniences of interpassive creation. He has confirmed that he wanted to give his original vocal a female tone, which he did by prompting the software to use “soulful vocal samples” (Savage, 2025). This is what Salvaggio (2025) similarly expresses when saying that technology prevents us from doing the hard, reflective work that real artistic growth requires to make the artwork meaningful. Rosalía resists this by cultivating a sound unlike anything typical in the mainstream charts, by blending classical orchestration with thirteen different languages. She even describes putting herself through the interactive process in the interview with Zane Lowe (Apple Music, 2025, 15:06): “How do I get from point A to point B. […] Instead of, from A to B, boom. […] No no no no no no. Go through the experience, as if the experience is… the most important is the journey. It’s not about getting from point A to B. Go through it. Let yourself go through it.”

Back to topIndustry survival

But why is all of this important? Even though we cannot categorize AI-generated art as meaningful, it is still doing well in the charts, and it will not go away anytime soon. Even though it plays directly into how Nguyen describes the current online ecosystem, it is disrupting the music industry.

First and foremost, AI models lack transparency and do not ethically source their datasets. Generative AI models are trained on vast unlicensed datasets scraped from the internet and music platforms. Existing copyright laws are ill-equipped to handle ownership disputes, leaving artists vulnerable (Berger, 2024). Again, we see this in the case of Haven. The AI-generated voice was so similar to singer Jorja Smith that her record label FAMM states that it believes the AI used Smith’s voice and is now seeking compensation. The AI platform Suno has admitted that its software was trained on copyrighted works (Savage, 2025). Because of the fast technological developments, AI is outpacing regulations. The UK is currently at the forefront of an AI copyright reform because policy makers are debating whether AI models are allowed to train on creative work without licensing (Berger, 2024).

This has devastating consequences for artists who are already dealing with streaming platforms like Spotify, which increasingly integrates AI-generated music to cut costs. This is while royalty streams get smaller for human artists. As Berger (2024) puts it: “Why license a costly catalog when AI can generate something “good enough” for free?” In response, musicians are pushing back. Over 1,000 artists, composers, conductors, singers, and producers contributed to the album ‘Is This What We Want?’, released in February 2025 as a protest against AI-generated music that fails to compensate human artists (Farley, 2025).

Beyond legal and economic concerns, artists also fear a cultural shift in which humans begin to sound like AI. Music has always been trend-driven: artists, fans, and executives chase what is popular, whether it be alternative rock, K-Pop, or Afrobeats. If AI-generated songs begin to dominate the charts, musicians may start echoing machine-made styles, leading to a downward spiral of homogenized “slop” (Farley, 2025). This threatens the core value of art and music since AI products lack lived experience.

Back to topConclusion

In conclusion, the comparison between Rosalía’s Lux and Haven’s I Run reveals that the value listeners assign to music is shaped not only by sound or genre but by the perceived integrity of the artistic process behind it. Rosalía’s Lux foregrounds human difficulty as a deliberate artistic choice that resists the acceleration and simplification of the attention economy in a linguistic, technical, and emotional form. Haven’s use of AI, by contrast, exposes how interpassivity can diminish the sense of meaning that arises from an artist’s intentional struggle with form, voice, and expression. The current music landscape shows that while AI can imitate style and produce commercially viable songs, it cannot recreate the human investment that gives art its purpose. Ultimately, difficulty becomes a kind of proof: evidence that a human mind and body engaged the world and shaped it into something new. As AI continues to alter the industry, this distinction may become one of the last remaining ways we recognize, value, and defend human artistry.

Back to topReferences

Apple Music (2025, November 5). ROSALÍA: The LUX Interview | The Zane Lowe Interview [Video]. YouTube. https://www.youtube.com/watch?v=jCcoi3Tzv_Y

Berger, V. (2024, December 30). AI’s Impact On Music In 2025: Licensing, Creativity And Industry Survival. Forbes. https://www.forbes.com/sites/virginieberger/2024/12/30/ais-impact-on-music-in-2025-licensing-creativity-and-industry-survival/

Bhandari, R. (2025, December 4). In the age of AI-everything, be Rosalía’s ‘LUX.’ The Nod Magazine. https://thenodmag.com/content/spanish-singer-rosalia-lux-new-album-13-languages

Coscarelli, J., & Caramanica, J. (2025, October 30). Rosalía’s Thrilling New Avant-Pop Swerve: Singing in 13 Languages. The New York Times. https://www.nytimes.com/2025/10/30/arts/music/rosalia-lux-interview.html

Farley, C. (2025, December 3). It’s Time to Rage Against the AI Music Machine. TIME. https://time.com/7338205/rage-against-ai-generated-music/

Kraaijeveld, S.R. (2024) AI-generated art and fiction: signifying everything, meaning nothing? AI & Soc 40, 217–219. https://doi.org/10.1007/s00146-023-01829-4

Nguyen, C. (2025, November 3). Is the Internet Making Culture Worse? Asterisk Magazine. https://asteriskmag.substack.com/p/is-the-internet-making-culture-worse

Petridis, A. (2025, November 3). Rosalía: Lux review – a demanding, distinctive clash of classical and chaos that couldn’t be by anyone else. The Guardian. https://www.theguardian.com/music/2025/nov/03/rosalia-lux-review-bjork-london-symphony-orchestra

Salvaggio, E. (2025, November 30). From interactive to interpassive. Cybernetic Forests. https://mail.cyberneticforests.com/from-interactive-to-interpassive/

Savage, M. (2025, December 1). Jorja Smith’s record label wants royalties from “AI clone” song. BBC. https://www.bbc.com/news/articles/cwyvjye18e9o

Stamp, T. (2025, November 15). Rosalia’s new album is everything pop usually isn’t (and why that’s a good thing). RNZ. https://www.rnz.co.nz/life/music/rosalia-s-new-album-is-everything-pop-usually-isn-t-and-why-that-s-a-good-thing

Steiner, G. (1978). On Difficulty. The Journal of Aesthetics and Art Criticism, 36(3), 263–276. https://doi.org/10.2307/430437

Zellner, X. (2025, November 4). How many AI artists have debuted on Billboard’s charts? Billboard. https://www.billboard.com/lists/ai-artists-on-billboard-charts/

Back to top