Since the emergence of basic AI song generators in 2022, significant progress has been made in the field. Presently, AI tools such as Suno.ai allow for the transformation of any set of words into musical phrases, even incorporating inside jokes, as evidenced by a recent tweet from the efficient engineer Riley Goodside. The tweet, posted on Wednesday, featured the intriguing statement “unhappy lady with piano performs the text of the MIT License,” which swiftly garnered attention within the IoT community.
The MIT License, a well-known liberal application license originating in the late 1980s and commonly utilized in open-source projects, was the focal point of Goodside’s tweet. The tweet’s opening and closing phrases, “my favorite part of this is ‘1:25 it nails’ Guarantees OF MERCHANTABILITY,” were accompanied by a beautiful Imogen Heap-style glissando and an interesting pronunciation of “FITNESS” as “fistiff.”
The inception of Suno, which translates to “listen” in Hindi, took place in 2023 in Cambridge, Massachusetts. The brainchild of Michael Shulman, Georg Kucsko, Martin Camacho, and Keenan Freyberg, individuals with prior experience at prominent companies like Meta and TikTok, Suno has garnered attention from major players such as Microsoft. Notably, Microsoft integrated an earlier version of the Suno engine into Bing Chat in December of the previous year. Suno, now in its v3 iteration, is capable of dynamically generating concise two-minute songs across various unique genres.
Despite our attempts to secure an interview, Suno had not responded at the time of publication. As reported by Brian Hiatt of Rolling Stone, Suno’s music generation model is described as a collaboration between Suno and OpenAI’s ChatGPT for lyric writing. However, concerns have been raised by experts regarding the potential use of copyrighted music recordings in the training of these models without appropriate artist or licensing agreements.
In a recent development, over 200 music artists signed an open letter urging the Artist Rights Alliance to refrain from using artificial intelligence in music creation processes that could potentially displace individual artists.
The issue of ownership surrounding AI-generated songs remains complex, particularly due to uncertainties regarding the training data’s origins. Suno’s FAQ states that songs created using its free tier are intended for non-commercial use and are owned by Suno. On the other hand, paying subscribers under the Pro or Premier tiers purportedly own the generated songs during their subscription period. While the legality of AI-generated music reproduction is yet to be definitively resolved, the US Copyright Office’s stance on AI-generated visual art may set a precedent for music as well in the future.
Suno’s capabilities in crafting unique songs based on user input or predefined themes are noteworthy, although the generated output may exhibit some imperfections. Despite the ongoing advancements in AI technology, there remains a discernible gap between AI-generated music and high-fidelity human-created compositions.
Our experiment with Suno involved requesting songs related to warriors with CRTs and incorporating two inside jokes, revealing an intriguing pattern where both songs commenced with the phrase “In the depths of the digital domain.” This consistency could indicate a hidden prompt guiding ChatGPT in the songwriting process.
While Suno undeniably serves as an entertaining tool in the realm of relational AI music, it is not devoid of ethical dilemmas surrounding the potential appropriation of artistic works without consent. Moreover, the prospect of AI potentially replacing human musicians has sparked concerns among individuals sharing their Suno-generated results online. Notably, AI influencer Ethan Mollick pondered the implications, stating, “I’ve had a track from Suno AI stuck in my mind all time. Great or terrible step?”
