AI
Striking a Discordant Note: The Ethical AI Music Generator That Fails to Harmonize
To go back to this article, navigate to My Profile, and then click on View saved stories.
The Newly Launched 'Moral' AI-Driven Music Creator Struggles to Produce a Good Tune
The launch of Jen, the new artificial intelligence music creator branded as "ethically trained," coincided with a notable moment. This was just after the Recording Industry Association of America had initiated legal actions against Udio and Suno for copyright infringement, accusing these platforms of using copyrighted content without authorization. Unlike its predecessors, Jen boasted about having legally obtained licenses for over 40 training collections and committed to rigorously screening all inputs and outputs to prevent copyright issues. As competing AI music generators were making headlines for all the wrong reasons, Jen was introduced as a more responsible choice.
Setting aside moral considerations, the quality of the music produced by Jen is a matter of contention. The tracks it creates lack vocals, and although it was trained using licensed content whose specifics have not been disclosed, it's unlikely to resemble anything familiar from mainstream radio. Requesting a song in the style of Willie Nelson from Jen is futile, as the program doesn't have the capability to replicate his sound. (When we tried, the result was closer to trip-hop than anything recognizable.) Even when specifying a broad genre like "outlaw country," which has a rich history and has shaped many iconic albums, the outcome is a generic form of "country" music that leans more towards easy listening than the gritty essence of honky-tonk.
Delving further into the matter, WIRED approached five seasoned musicians to evaluate Jen's capabilities: John Heywood, a renowned bass player famous for his performances with the indie rock band Alex G; Jenn Wasner from Wye Oak and Flock of Dimes; Shana Cleveland, the creative force behind the surf-noir band La Luz; Steve Reidell, known for his work with The Hood Internet and Air Credits; and Allen Blickle, a composer and sound designer who has been nominated for two Emmy Awards and has contributed to projects for major companies like Netflix, Disney, and Apple Music. Each of the musicians found the software straightforward to use, yet lacking in inspiration.
Wasner has expressed openness to leveraging AI as a creative brainstorming tool, yet her firsthand experiences, particularly with Jen, left her feeling unsatisfied. Despite witnessing AI's application in music production and songwriting, she found the output disconcertingly artificial. To her, the creations seemed to stem from an eerie, almost unnatural origin, intriguing yet ultimately hollow. "Listening to the AI's compositions was like witnessing a parlor trick, such as unexpectedly mixing genres like adding electronic beats to traditional music," she described. For Wasner, the AI-generated ideas never truly resonated or sparked genuine excitement. She often reflected that her own creativity could surpass what the AI proposed, lamenting, "I always believed I could devise something far more interesting myself."
Certainly, our evaluators took their inquiries with Jen to extremes not typically expected from the average user, opting for a depth of knowledge akin to what one might expect from a seasoned record store employee. For example, Cleveland was unsuccessful in extracting a satisfactory response to a request for "mid-tempo California garage rock with a '70s Indonesian pop influence," while Heywood was disappointed that Jen couldn't identify his mention of "city pop," a genre of Japanese music that gained traction in the mid-1970s and has recently experienced a slight revival in interest. For Heywood, having access to such a wide-ranging musical knowledge is crucial, particularly in his role as a musician.
"At no time did I consider the notion appealing, always feeling confident that I could have devised a more interesting concept myself."
Heywood points out that when musicians or producers request something from one another, they often mention other bands or artists for context, saying things like, ‘We’re aiming for a sound reminiscent of Prince,’ or, ‘Incorporate a Clavinet similar to Stevie Wonder's style.’ However, Jen’s unfamiliarity with current recording artists, as well as with certain well-known genres and instruments, complicates the process of achieving a precise sound.
Heywood was attempting to extract some kind of warmth from it, aiming for a sound that was either lo-fi, vintage, or had a vinyl crackle or a rich saturation, but all his efforts resulted in a sound that was consistently hi-fi, reminiscent of a video game menu. He mentioned that even when using 'lo-fi' as a suggested prompt, it barely made a difference. Heywood highlighted that if one's goal is to achieve a specific audio aesthetic, such as the vibe of '80s funk, the outcome tends to lean more towards a sound akin to Daft Punk rather than hitting the desired mark.
The electric guitar tones produced by WIRED and its evaluators were consistently too pristine, and achieving a track in any time signature other than 4/4 seemed nearly unattainable unless the term “waltz” was included in the instructions.
According to Shara Senderoff, a cofounder of Jen, the current limitations of the tool are part of its early development stage. Currently in alpha, the tool produces tracks of 10 and 45 seconds in length that are designed to spark creativity and act as a preliminary basis for further creative work, rather than serve as complete works in themselves, she explains. Senderoff also notes that the tool, which was developed with a constrained set of data, is poised for substantial growth and enhancement during its upcoming beta phase, promising new functionalities.
Heywood comments that whatever Jen created, claiming it to be rock music, resembled a superficial imitation of the genre, similar to "clip art." According to Cleveland, Jen managed to produce a few tracks that seemed suitable for a car advertisement or reminiscent of The Black Keys' style. However, Cleveland primarily found Jen's musical ideas to be overly clichéd and uninspired.
"She described the sound as something she might create while having fun with pals, playfully teasing the tropes of different music styles," she mentions. "She imagined a few tracks might fit into a poorly rated Netflix romance series, yet none of the creations seemed to challenge her own identity."
However, what about those behind the scenes creating the music you might encounter on a Netflix romance program? Is there a possibility that Jen could jeopardize their employment? Blickle believes that it's highly probable.
"For small-budget producers eager to distribute their work, the option to bypass hiring a designer or animator by utilizing an image generator is now on the table," he explains. "This holds true for music production costs as well. If what was previously a $2,000 expense can now be avoided entirely, then it's seen as a direct saving of $2,000 for someone."
Applications such as Jen currently lack features that allow users to designate climax moments or incorporate brief, impactful musical phrases into their compositions, features that would render these tracks instantaneously suitable for soundtracks. However, it's reasonable to anticipate that these capabilities will be introduced in the near future.
Blickle speculates that repositories of stock music, frequently utilized by smaller productions and reality shows for their scoring needs, are likely to experience a surge in contributions from entities akin to Jen. Although he labels much of the content in these libraries as "garbage," he notes that "for those in search of such music, the quantity available is only going to increase. This will lead to a vicious cycle of subpar library music, and I fail to see how this enhances anyone's quality of life."
This might significantly clutter the offerings on digital music platforms such as Spotify, which already sees an influx of thousands of new tracks daily. Blickle notes, “The volume could increase two to fivefold due to individuals utilizing AI aiming for 100,000 listens on one track, or perhaps a million on another. Their goal is simply to put out music and earn some additional income, and that poses a concern.”
The surge in music produced by artificial intelligence might lead to complications for productions opting to incorporate AI music. Platforms such as YouTube and Instagram may permit the uploading of songs crafted by AI, under the premise that you possess ownership of this "original" piece. However, regulations concerning AI-generated content differ globally, indicating that if you aim to distribute a TV show internationally, caution is advised regarding the origins of your music—even if it originates from a "responsible" entity like Jen. Additionally, in the United States, the law requires that a human creator is associated with any work that seeks to obtain copyright protection.
A quarter-century on from the Napster era, all five musicians expressed a sense of resignation over the potential impact of music-creating technologies such as Jen—regardless of their ethical implications—on their careers. “The avenues for musicians to earn from their craft are dwindling, and the prospect of losing yet another source of income is certainly disturbing. However, I've come to terms with this reality,” stated Wasner. “It’s something we're going to have to adapt to.”
Recommended for You…
Direct to your email: Dive into the future with Will Knight's Fast Forward, your guide to the latest in AI advancements
Delving into the largest undercover operation ever conducted by the FBI
The WIRED Artificial Intelligence Elections Initiative: Monitoring over 60 worldwide electoral events
Ecuador finds itself completely without recourse amid a severe drought.
Be confident: Here are the top mattresses available for purchase online
Additional Content from WIRED
Critiques and Manuals
© 2024 Condé Nast. All rights are protected. WIRED might receive a share of revenue from products bought via our website, as a result of our affiliate agreements with retail partners. The content on this website is not allowed to be copied, shared, broadcasted, stored, or used in any form without the explicit written consent of Condé Nast. Choices in Advertising.
Choose a global website
Discover more from Automobilnews News - The first AI News Portal world wide
Subscribe to get the latest posts sent to your email.