Copyright and AI – What you need to know!

Ed Sheeran – uncrowned King of copyright lawsuits. In recent years, the singer has had to defend himself so often against accusations of plagiarism that yet another lawsuit hardly seems worth a headline. When it comes to the lawsuit started against him by Edward Townsend, however, things are different. Townsend had noted striking similarities between the classic "Let's Get it On", which Townsend had written with Marvin Gaye in 1973, and Sheeran's "Thinking Out Loud". After four tough years, the court ruled in Sheeran's favour, causing some in the alternative music press to rejoice. As online magazine Pitchfork put it: "Ed Sheeran Finally Seems Useful. It took a serial plagiarist to uphold creative freedom." Yet the result hardly clears things up. And in the long run, it could rather pave the way for a final departure from the current copyright system than its salvation.

For one thing, the decision illustrates how capriciously legal principles are currently interpreted. Only a few years ago, Californian judge John A. Kronstadt had pioneered the position, in the case of "Blurred Lines" written by Robin Thicke and Pharell, that it was too similar to the feel and sound of (again) Marvin Gaye's "Got to Give It Up". The duo had to pay back millions in royalties - a precedent with still potentially disastrous consequences. Even an open letter signed by 200 well-known musicians failed to sway the jurors' minds at the time. Little wonder many felt satisfaction at the about-face. And admittedly, it does offer hope that the rights of active songwriters could be given more weight again.

Ironically, "Thinking Out Loud" and "Let's Get it On" have much more in common than just their chord progression. If you listen to the two songs in the background, you could be forgiven for thinking the same piece were playing twice in a row. From groove and tempo to accentuation and instrumentation to sound - the producer team around Jake Gosling expertly made the result sound like a karaoke version of Gaye's classic. In an interesting video, producer Rick Beato revealed these clear references at the beginning of the process. At the same time, he also pointed out that there are a lot of differences: The lyrics and lyrical themes, the harmonic development and above all the melody. It simply depends on where your priorities lie.

So what, exactly, defines "similarity" in a musical sense?

One would think that such a fundamental question should have a simple answer. In fact, it has provided lawyers and courts with a secure income ever since the earliest years of the music industry - that is, since Handel and Mozart. Generally speaking, only melody and lyrics can be protected, while the rhythmic aspects of a track are considered fair game for anyone. But in a broader sense, the idea of the system is to protect all the elements that make a song unique and recognisable. Even a chord progression could thus be covered by copyright, if only it is original enough. This approach seems to make a lot of sense. But, really, it's little more than a tautology: what makes a song original is its originality. No wonder there have been countless hotly contested cases over the last 100 years.

The support for Sheeran is easy to understand. As the number of lawsuits continues to increase, the situation for living musicians is becoming increasingly tricky. Pop music, by default, makes use of just a few building blocks and originality is hardly its main concern. What's more, songwriting teams are growing and, as a result, more and more influences are flowing into every production. If you have committees made up of up to 20 members, some of whom contribute only a small building block to the final product, it becomes almost impossible to guard yourself against possible "contamination". Labels, on the other hand, are virtually clamouring for carbon copies of hits so tracks end up in popular playlists on streaming services and are picked up by algorithmic recommendation tools. Only very few manage this balancing act: Too close to the original and you're labelled a thief. Too far from it and the chances of success dwindle.

Who knows whom?

As if this were not difficult enough, a crucial question is whether the alleged plagiarist knew the "original" or not. Even if a song is exactly the same as an existing one, it is not considered plagiarism if the similarity is purely coincidental. George Harrison's "My Sweet Lord" was obviously identical to the Chiffons hit "He's so fine", but Harrison claimed he had never heard the song. The Chiffons had scored a big hit with it, however, and the presiding judge simply thought it more likely that Harrison had unconsciously heard the song and equally unconsciously recreated it - Harrison had to pay up. An earlier Ed Sheeran case was different: Sami Chokri and Ross O'Donoghue claimed that part of the mega-hit "Shape of You" resembled their "Oh Why". Here, however, the judge took the opposite stance: Certain passages of the two songs were indeed similar. But it was unlikely that Sheeran had lifted them off "Oh Why" because the song was simply too unknown. The plaintiffs came away empty-handed.

It doesn't take a degree in rocket science to understand that this ruling is problematic. It leads to a world in which the interests of established, even long-dead artists outweigh those of a new generation. It virtually implies that stars have earned the right to freely help themselves to the work of unknown colleagues, while the latter must carefully scrutinise every note. That aside, the music world has changed radically since the Harrison verdict. As radio and television lose ground, streaming figures will be the main indicator of a song's popularity in the future. But how many streams does one need for a song to be considered big enough to assume everyone has heard it at least once?

Perhaps it is finally time to rethink existing copyright.

While musicologists argue about melodic arcs and the uniqueness of harmonic sequences, similarity is being completely redefined by new technological tools. Streaming providers (like Spotify and Apple Music) have relied on recommendation algorithms as a customer retention strategy. Whether Edward Townsend or Ed Sheeran gets a royalty cheque is irrelevant to these services. What matters to them is the time users spend on their apps and the monthly fee that comes with it. The best user experience is a playlist in which the listener doesn't skip a song because each new track perfectly satisfies her taste. Curiously, many of the criteria that are currently so important in the court room hardly play a role here. What matters far more are feeling, mood, sound, groove and rhythm, scene affiliation - all that makes for the striking similarity between "Thinking about you" and "Let's Get it On".

"Too close to the original and you're labelled a thief. Too far from it and the chances of success dwindle.No wonder there have been countless hotly contested cases over the last 100 years."

There are a variety of ways to realise this goal. One of the most interesting new approaches is to convert a track into an image that is much deeper than a mere waveform and to then compare it with the images of other compositions. This approach makes use of extremely advanced image analysis techniques and brings to light patterns that defy traditional evaluation. Most recommendation algorithms, however, simply make use of brutal computing power. They work with huge amounts of data applied to an extremely long list of descriptive tags. They also take into consideration correlations from social media: if two songs appear frequently in the same networks on twitter, they are considered more closely related. Language and cultural groups can also be included using appropriate statistical methods. This creates gigantic trees of similarity and differences.

All this may seem rather strange in view of the emotionality of music. But the results prove the inventors right: automated recommendations are largely on par with or even superior to curated playlists created by human experts. And compared to traditional approaches, these methods have another advantage: they are far more objective and allow for verifiable judgements. They will certainly not be used in court any time soon. But that may not be necessary. If they succeed in expanding the criteria for originality and putting them on a more solid foundation, this alone would help all those who are looking for something personal in their art instead of just using the hits of the moment for “inspiration”. That said, Ed Sheeran might be in need of a new songwriting strategy.