The regulation of AI in the arts is a hot topic. The music industry is unique as several music labels control the licenses to most of the music that has been produced in the past half a century.
Most, but not all.
Derivative music is created using material to make a new composition or sound recording from a pre-existing work. This has created a gray area because the input, or original composition, is also usually protected by copyright. AI wants high quality music because the ultimate quality of output is heavily dependent on the quality of input. Unhappy with “the training of generative AI using our artists' music”, in April 2023, Universal Music Group invoked copyright violation to take down the track “Heart on My Sleeve” allegedly written by AI to sound like it was by Drake and The Weeknd.
The aforementioned problem of input legitimacy is relevant for images and texts that are applicable for AI. The core difference lies in the accessibility of input data on which to train AI. MidJourney and GPT were trained on images and texts they could mostly use without the consent of their respective creators. However, some copyrights may have been violated. The company Stability AI recently got into litigations with stock photo provider Getty Images, which wants to prevent selling its AI image-generation system in the UK and US. In late 2022, three artists formed a case to sue multiple generative AI platforms on the basis of AI using their original works.
Is there an argument for the idea that, since we pass the input through the neural black box, it is possible that the result will not even resemble the input and thus be free of infringement charges? Not really.
AI has exacerbated a legal framework that barely worked even decades ago. One of the extreme cases relevant to the current topic is that of rapper Vanilla Ice’s legal issues with the band Queen and artist David Bowie. The hook in the superhit “Ice Ice baby” (1990) bore some resemblance to Queen’s “Under Pressure”, but was not completely the same; Vanilla did add one extra note. This was a wise move and he could have proven his authorship in the courthouse. However, the artist was quick to pay $4 million for the track. This is because the lawsuit about whether the additional note makes him free of copyright infringement could have cost even more.
Later he explained that sampling is a state of mind, which is true. Rap music makes the clearance of rights a fertile ground in the music business. However, generative AI has lowered the entry barrier for sampling. Therefore, the thousands of tunes it would be possible to produce in the blink of an eye would require, accordingly, thousands of new clearance offices. These would undoubtedly have plenty of work because generative music has recently uncovered a use, which is very interesting, albeit difficult to enjoy.
That is the robust pace of the Creative Economy, for example bloggers, streamers, and so on. They require music to accompany their content, which has to be generated on demand by a set of parameters. These, in turn, have to be rewarded. They are used for the needs of output content that is placed on platforms with relatively firm and rigid rules on copyright.
This opens up the opportunity for human musicians to contribute to AI with their own music. Writing music across all genres, pitches, and moods that would legally enter AI’s black box is a job musicians can do to support their respective musical endeavors. Сashflow generated by the Creative Economy’s consumption of legal AI works may help support families (I know some real stories about that) and allow some bedroom musicians to enter the professional arena.
Conversely, attempts to measure human involvement in the creation of the final works may entail endless bureaucracy. This is also an impractical and irrational approach because it involves trying to find and prove the human touch in something created by a machine. At least, this is what we tell ourselves. However, paradoxically, we are giving the machines a considerable ascendance. This is because, if there were a proven portion of human touch in the final works, it is likely that the machines would like to ask the humans: but who created the rest? The machine would be a full-scale contributor to the musical work and its legitimate co-author.
The most practical approach is to keep in mind that AI is nothing more than a tool for humans to use to benefit industry and society. The best and only way to value the human touch is to avoid any unlicensed content to input generative AI. This tool will undoubtedly be a benefit to the creative economy but the question remains as to whether the same will be able to be said for human artists.
Ironically, artists have more opportunities to benefit in Europe because European regulations are far more fierce and restrictive. Previously, this approach bore little fruit. However, it may now benefit musicians by generating cash flow from royalties for AI’s input.
So, the future of the whole emerging industry depends on our attitude to the AI black box; do we consider it to be a co-author and try to evaluate its contribution in the final works, or do we use it as a useful tool and feed it with licensed input?
Vanilla Ice preferred to license the input of his black box. It now doesn’t matter whether it was Vanilla Ice or Queen who wrote the simple yet genius bass riff, or whether one additional note solved the issue. It doesn’t matter because both versions now belong to Vanilla Ice, in a deal he dubbed the best ever.
The post Artists Behind Neural Network Models: The Impact of AI on the Creator Economy appeared first on Unite.AI.