Lentis/AI Music, Creativity, and Intellectual Property

Overview

Introduction

Artificial intelligence (AI) is revolutionizing the music industry by redefining how music is composed, performed, and consumed. What began as theoretical musings by pioneers like Alan Turing has grown into a field with sophisticated neural networks capable of composing full-length songs. However, this evolution raises pressing questions about originality, ownership, and the future of human creativity.

History

Theoretical Foundations: 1950s-1960s

The concept of AI-generated music originated with Alan Turing, who theorized in the 1950s that machines could emulate human cognitive processes, including music composition. His ideas laid the groundwork for computational creativity.

In 1957, Lejaren Hiller and Leonard Isaacson at the University of Illinois created the Illiac Suite. Using the ILLIAC I computer, they developed an algorithmic process to compose a string quartet, marking one of the first successful instances of computer-generated music. It demonstrated the potential of computers to collaborate in creative endeavors, sparking decades of research into AI's role in music.

Rule-Based Systems: 1970s-1980s

During the 1970s and 1980s, AI systems advanced with rule-based algorithms that mimicked specific musical styles. David Cope’s Experiments in Musical Intelligence (EMI) exemplified this approach by analyzing existing compositions and generating new ones in the style of Bach and Mozart. However, EMI heavily relied on predefined human rules, highlighting the imitative, rather than generative, nature of early AI.

A significant technological leap came in 1983 with the introduction of MIDI (Musical Instrument Digital Interface). MIDI standardized communication between computers and electronic instruments, enabling precise control of musical elements and facilitating data collection for future AI systems.

Machine Learning: 1990s-2000s

AI music systems transitioned from rule-based models to data-driven techniques in the 1990s. Machine learning algorithms enabled pattern recognition and predictive modeling, allowing AI to generate more complex and varied compositions. Projects like Accenture’s Symphonologie showcased AI as a collaborative tool, blending human and machine creativity.

Neural Networks: 2010s-Present

The rise of deep learning revolutionized AI in music. Neural networks, trained on vast datasets, now compose intricate, multi-instrumental pieces in seconds. Tools like OpenAI’s MuseNet and AIVA exemplify how AI has expanded its creative capabilities across genres, challenging the boundaries of human artistry.

Relevant Legal Disputes

Burrow-Giles Lithographic Company v. Sarony, 111 U.S. 53 (1884)

Oscar Wilde No.18 by Napoleon Sarony (1882)

Burrow-Giles Lithography Co v. Sarony was submitted on December 13, 1883 when Sarony filed a copyright suit against Burrow-Giles when the company had marketed lithographs of Sarony's photograph, Oscar Wilde No.18, without Sarony's permission. Burrow-Giles argued that photographs were neither writings nor a production of an author, and thus did not possess copyright protection. The court determined that even though creating a photograph is "merely mechanical, with no place for novelty, invention, or originality", the photograph Sarony created was his "own original mental conception, to which he gave visible form by posing the said Oscar Wilde in front of the camera, selecting and arranging the costume, draperies, . . . suggesting and evoking the desired expression, and from such . . . representation, made entirely by plaintiff, he produced the picture in suit.'" On March 17, 1884, the court established that photography is a copyrightable medium. Furthermore, this case reinforced that an author of a work is "he to whom anything owes its origin."

Naruto v. Slater, No. 16-15469 (9th Cir. 2018)

One of the monkey selfies

Naruto v. Slater, more commonly known as the Monkey Selfie copyright dispute, began on September 22, 2015. PETA, representing the crested macaque Naruto, sued Slater claiming that because the monkey took its own picture, it was entitled to the copyright. David J. Slater, the photographer, argued that he held the copyright as he engineered the situation that led to the monkey taking its selfie. On April 23, 2018, the court determined that "this monkey - and all animals, since they are not human – [lack] statutory standing under the Copyright Act", reinforcing that non-humans cannot own copyright.

Thaler v. Perlmutter, No. 22-1564 (BAH)(DDC 2023)

A Recent Entrance to Paradise

Thaler v. Perlmutter emerged after the U.S. Copyright Office denied Dr. Stephen Thaler's copyright application for an AI generated work, "A Recent Entrance to Paradise". The court focused on answering "whether a work generated autonomously by a computer falls under the protection of copyright law upon its creation". Thaler filed a lawsuit, arguing that the copyright of the AI generated work belonged to him as the AI model's creator. He stated that "AI should be ‘acknowledge[d] . . . as an author where it otherwise meets authorship criteria, with any copyright ownership vesting in the AI’s owner.’" The court reviewed the decision of Burrow-Giles Lithographic Co. v. Sarony, finding that the author of any AI generated work would be the AI. However, following the decision of Naruto v. Slater, since AIs are not human, they cannot hold copyright. The case was decided on August 18, 2023, confirming that AIs cannot hold copyright. As such, the work fell under public domain. However, the court explicitly stated that they were not making any decisions on "how much human input is necessary to qualify the user of an AI system as an 'author' of a generated work".

Zarya of the Dawn

Zarya of the Dawn by Kristina Kashtanova

Zarya of the Dawn, a comic book authored by Kristina Kashtanova and illustrated using the AI generator Midjourney sparked a significant copyright dispute in 2022. When Kashtanova applied for copyright of the comic in September 2022, she failed to disclose that the images were generated by Midjourney. The comic was originally granted copyright by the Copyright Office, but after discovering Midjourney's involvement in illustrating the comic, they reviewed their decision. While the Copyright Office agreed that the text and compilation (the arrangement of images and text) were copyrightable, the art was not, as the "process is not controlled by the user because it is not possible to predict what Midjourney will create ahead of time." Although, Kashtanova claimed "she personally edited some of the images created by Midjourney", the Copyright office could not determine "what expression in the image was contributed through her use of Photoshop as opposed to generated by Midjourney" and thus revoked the copyright.

Relevant Copyright Chapters

Chapter 308: The Originality Requirement

The Originality Requirement has two primary requirements: "The work must be ‘independently created by the author’ and it must possess ‘at least some minimal degree of creativity’"

Chapter 308.1

The Copyright Act defines "independent creation" as a work created "without copying from other works." However, "a work may satisfy the independent creation requirement ‘even though it closely resembles other works so long as the similarity is fortuitous, not the result of copying.’"

Chapter 308.2

In terms of creativity, the Copyright Act states that "‘the requisite level of creativity is extremely low.’ Even a ‘slight amount’ of creative expression will suffice.’"

Works That Do Not Satisfy the Originality Requirement

313.4(A): Mere Copies

313.4(K): Mere Variations of Colorings

Chapter 306: The Human Authorship Requirement

The Human Authorship Requirement states that only humans may own a copyright. "The U.S. Copyright Office will register an original work of authorship, provided that the work was created by a human being."

Chapter 313.2

The chapter states that "the U.S. Copyright Office will not register works produced by nature, animals, or plants." Furthermore, "the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author."

Works That Lack Human Authorship

• A photograph taken by a monkey.
• A mural painted by an elephant.
• A claim based on the appearance of actual animal skin.
• A claim based on driftwood that has been shaped and smoothed by the ocean.
• A claim based on cut marks, defects, and other qualities found in natural stone.
• An application for a song naming the Holy Spirit as the author of the work.

Chapter 507.1: Derivate Works

As AI generated works are considered to be authored by the AI, and thus public domain, attempts to edit and then copyright the works would most likely fall under derivative works. The Copyright Act defines a derivative work as "a work based upon one or more preexisting works, such as a . . . art reproduction . . . or any other form in which a work may be recast, transformed, or adapted."

Chapter 311.2

Derivative works follow the same originality requirements as original works. Derivative works may only receive copyright protection "provided that [the work] contains a sufficient amount of original expression, meaning that the derivative work must be independently created and it must possess more than a modicum of creativity" The Copyright Act does reiterate that "[m]iniscule variations do not satisfy this requirement, such as merely changing the size of the preexisting work."

Present Issues

Lack of Creativity

In 2023, an anonymous creator known as "Ghostwriter" released Heart on My Sleeve, featuring AI-generated vocals mimicking Drake and The Weeknd. The song went viral, drawing millions of streams before being removed for copyright infringement. Universal Music Group argued that the AI-generated track exploited the artists’ likenesses without consent, raising ethical and legal concerns.

This case highlights AI’s ability to imitate human creativity so convincingly that it blurs the line between originality and replication. Although the creator faced no legal repercussions, the controversy underscored the need for clearer regulations on AI-generated content.

• AI doesn’t create entirely new works; it amplifies patterns derived from training data. While it can replicate styles, it lacks the emotional depth and intentionality of human artistry.
• Tools like MuseNet and Suno allow anyone to compose music in minutes, potentially devaluing human creativity by making artistic production instantaneous and impersonal.

Suno and Udio

The most current case reflecting the debate regarding the originality of AI music is in the RIAA vs. Suno and RIAA vs. Udio. The RIAA, or the Recording Industry Association of America, is a trade organization that represents many record labels, including Sony Music Group, Universal Music Group, and Warner Music Group (colloquially known as "the big Three" in the industry as they own up to 85% of market share). In June 2024, the RIAA placed a copyright infringement lawsuit on Suno and Udio (two generative AI music startup companies). They had three main goals with this lawsuit:

1. To have these companies admit their models use copyrighted music.
2. Immediately stop training their AI models using this copyrighted music.
3. Seek damages up to $150,000 per song used in the models.

When first discussing the lawsuit with the press, Chairman and CEO of RIAA, Mitch Glazier, stated:

"The music community has embraced AI and we are already partnering and collaborating with responsible developers to build sustainable AI tools centered on human creativity that put artists and songwriters in charge, but we can only succeed if developers are willing to work together with us. Unlicensed services like Suno and Udio that claim it's 'fair' to copy an artist's life's work and exploit it for their own profit without consent or pay set back the promise of genuinely innovative AI for us all."

The RIAA hopes to reflect that the inherent moral value of this statement for pursuing this lawsuit is protecting artists' work and fair pay. However, in any of RIAA's arguments, the percentage of and how the money collected from damages will be distributed to the artists of each respective record label must be clarified.

Suno responded to the copyright lawsuit, "What the major record labels really don't want is competition." While this calls to question the actual motive of the lawsuit and whether the benefactors will include artists, the results of this lawsuit are pertinent as they will set further precedents in how copyright pertains to AI overall.

Initially, in response to the lawsuit, Udio claimed "fair use," which is primarily used for academic or journalistic purposes. However, lawyers dismissed this claim because their training is not used for human-created content.

As the lawsuit proceeded, both companies admitted to using copyrighted music only as "intermediate copies." They sighted a copyright case from 2021 regarding an AI algorithm's use of copyrighted images to create a search algorithm. In that case, using these "intermediate copies was lawful" because these lower-resolution copies did not satisfy the user's desire to see the full images. However, in Suno and Udio's context, their AI models are not search algorithms but are generative. 

Other AI Music Lawsuits

Following this case, different record labels (primarily Universal) have filed a copyright on issues regarding AI music. On October 18, 2023, several record labels, namely Universal, filed a lawsuit alleging that Anthropic's AI chatbot, "Claude," unlawfully copied and disseminated copyrighted works, including lyrics "owned or controlled by the Publishers.

On January 31, 2024, the copyright license agreement between Universal Music Group (UMG), the world's largest music company, and TikTok owner ByteDance Ltd. expired without a renewal agreement. TikTok was forced to turn off the reproduction of all audio content from artists belonging to UMG's catalog, including Taylor Swift, Billie Eilish, Romeo Santos, J Balvin, and Imagine Dragons, among many others. During these negotiations, UMG made AI-generated music a sticking point in their TikTok negotiations, and a group of musicians spoke out against AI-generated music. Groups, including the Recording Academy, the Music Workers Alliance, the National Association of Music Publishers, the American Association of Independent Music, and even SAG-AFTRA, have all made statements supporting the lawsuits.

Ethical Implications of AI in the Music Industry

As is apparent with the growing number of lawsuits directed at AI music companies, advancements in AI bring new possibilities for creativity and efficiency along with critical ethical concerns regarding economic impact, bias, and data harvesting.

AI’s role in the music industry is a double-edged sword, offering both opportunities and challenges. On the one hand, AI automates many aspects of music creation and production. Tasks like composition, audio mastering, and even lyric writing can now be performed by AI tools. However, this automation has led to concerns about job displacement. Traditional roles in the industry, such as sound engineers and live musicians, face increasing obsolescence as AI tools become more sophisticated and accessible. However, AI also democratizes music production, enabling independent artists and smaller labels to create high-quality music without needing expensive equipment or studio time. So, while jobs are being lost, new roles are emerging, such as AI music curators and prompt engineers. This demonstrates how technological advancements can create fresh opportunities even as they disrupt existing systems.

Another critical issue in the use of AI in music is the presence of bias in the models used to generate content. AI systems are only as good as the data they are trained on, and this dependency can result in significant imbalances. For instance, if an AI model’s training data predominantly features Western music, the system may struggle to generate accurate representations of non-Western genres or cultural styles. Along with racial bias, gender bias in datasets used for AI training can perpetuate existing inequalities in the music industry. If female and non-binary artists are underrepresented in the data, AI-generated music may disproportionately reflect the styles and contributions of male artists. Such biases highlight the need for diverse and inclusive training datasets. Developers must address these issues to ensure that AI tools reflect a broader and more equitable spectrum of musical expression.

As discussed previously, the question of how AI systems source their training data has sparked widespread ethical debates. Many AI models, most notably Suno and Udio, are trained using copyrighted music, often without the consent or compensation of the original artists and have received backlash. This practice raises significant concerns about exploitation. Artists invest considerable time and effort into creating music, yet their work can be appropriated by AI systems without acknowledgment or financial remuneration. This mirrors controversies in other areas of AI, such as the exploitation of the unpaid labor of associated with Remotask workers. If creators cannot safeguard their work against unauthorized use by AI systems, the industry risks devaluing artistic contributions in favor of technological efficiency. The music industry is at a crossroads–do we protect the rights of creators, or do we embrace AI-drive efficiency without clear ethical guidelines?

Reception

Students

While students seem to find AI music fascinating, they emphasize that AI music lacks the emotional depth found in human compositions. They explain that human compositions are capable of conveying authentic human experiences and feelings, which AI cannot replicate.

Quotes are pulled directly from a New York Times article:

"Human singers have the ability to convey a wide range of emotions and add their own unique style to their music. They have the creativity and experience to interpret lyrics and to give them their own style. Music made with A.I. may be good, but it will never be the same as music sung by a good pop star. While A.I. can be a helpful tool for composers and producers to experiment with new sounds and styles, it will never replace the raw talent and artistry of human vocalists." - Sergi, Sant Gregori

"In my personal opinion, the main allure of music lies in the emotional connection that it establishes with the listener. It’s not only about the melody or the lyrics, but also the sentiments and reminiscences that the music evokes. If I became aware that a song was completely generated by an artificial intelligence system, I would feel intrigued to listen to it, driven by my fascination with the underlying technology. Nevertheless, I might not feel as connected to the song on an emotional level as I would if it was composed by a human artist." - Aleena, Julia R. Masterman, Philadelphia, PA

"Even if the lyrics emulate a real pop star’s lyrics, whether it talks about heartbreak, love, or family, it would never have the same impact like one created by a real pop star because the listeners know it is all just to sound good and appeal to the listeners. It would have nothing to do with the artist themselves … Music is supposed to be relatable. When I listen to music, I like to listen to artists I know have gone through similar things or have the same feelings as me." - Sophia, Hinsdale Central High School

Composers

The evolution of composers attitudes towards AI music is reflected in Daniel James, a composer on YouTube, perspective over time. While he acknowledged that AI generated music was inevitable, it would never allow a total newbie to suddenly become a professional. He suggested if AI became proficient in replicating professionals, composers would need to come up with their own style away from the AI. James had a positive view on AIs potential as a tool, comparing it to a ghost composer that could handle much of the monotonous work that an assistant might do. He predicted that AI would primarily impact the lower-skilled composers, widening the gap between low and high skilled composers.

By 2023, James' perspective has become more nuanced. He discussed the limitations of AI, explaining that it can't really create something new, and can only make variations. While acknowledging AIs capability to blend styles, like reimagining Beethoven as hip-hop, he emphasizes that music exists within the context of the time it is created. James considers that although AI can accurately replicate music, it still struggles with emotional nuance. He also notes that skilled AI use would still require musical expertise to properly guide the AI. James also discusses the significant impacts AI might have. Formulaic music like EDM and trailer music seem particularly vulnerable to AI as there is a "correct" way to write those styles of music. He also believes that AI music will severely damage the lower end of the creative field. He worries that if newer generations of artists are trained to use AI, they may forget how to develop their own original styles, stunting their growth.

The broader composer community has similar views. On Reddit, many composers advocate for limiting AI's role to reference and placeholder work that bridge the gap between the client and the composer. The consensus among professional composers, including James, is that AI needs much better regulation. They argue for protecting artists' unique, copyrightable styles from unauthorized AI replication and advocate that AI companies should compensate the artists whos works train the AI models. These composers are working to convince the Copyright Office that training AI on artists' work without compensation devalues artists.