Spotify just invented AI technology that will police songwriter plagiarism

Credit: Shutterstock

Songwriters of global hits getting sued for alleged plagiarism has become a recurrent story on MBW these past few years – and a recurrent source of misery for writers and their representatives in the industry.

But what if a songwriter or composer were able to use AI technology to avoid litigation altogether, by finding out if their song copies elements of other compositions, potentially in real time?

That could now be a reality, thanks to a Spotify invention revealed in a new European patent filing from the company obtained by MBW.

According to a document published last week, Daniel Ek‘s company is seeking a patent for its “Plagiarism Risk Detector And Interface” technology, which pertains to “Methods, systems and computer program products..for testing a lead sheet for plagiarism”.

As explained in the filing – and as our songwriter/musician readers will already know – a ‘lead sheet’ is a type of music score or musical notation for songs denoting their melody, chords and sometimes lyrics or additional notes.

Spotify’s invention would allow for a lead sheet to be fed through the platform’s ‘plagiarism detector’, which would then, “having been trained on a plurality of preexisting encoded lead sheets”, immediately compare the composition in question to all other songs stored in its database.

A set of messages would then be displayed – describing a detected level of plagiarism regarding “a plurality of elements” such as a chord sequence, melodic fragments, harmony, etc. of a song  (see fig 7 below).

The AI software would also potentially calculate “a similarity value” of the song in question vs. other songs in the Spotify lead sheet library.

These technology could work the other way around, too, says Spotify’s filing, reassuring a songwriter that “the melodic fragment [of your song] appears to be completely new”.

One particularly interesting element of this is that it would take place in near-real time, allowing a songwriter or composer to tweak elements of their work to avoid infringement before they (and/or their record label) spent the big bucks on recording a final version.

Spotify’s filing adds that “in some embodiments a link to the media content item that might be infringed (e.g., a track of an album) is provided so that a [songwriter] can quickly… listen to the potentially plagiarized work”.

Various other techniques and systems have been previously proposed for detecting plagiarism in musical compositions are referenced in the patent filing.

Some of those include the Music Plagiarism Detection System “that detects music plagiarism based on melodic similarity” as well as the Audio Forensics Meets Music Information Retrieval – A Toolbox For Inspection Of Music Plagiarism, which proposes “detecting and inspecting sampling plagiarism, rhythm plagiarism, and melody plagiarism”.

Spotify argues that while such techniques “are significant improvements over manual approaches, they still require significant expertise and are not suited for operation by typical artists and composers, especially artists and composers who are interested in detecting plagiarism during the composition process”.

The company’s solution to this problem “is a graphical user interface (GUI) that is more intuitive, more precise as to the portion of the work that may be considered plagiaristic, and that provides dynamic visual feedback in substantially real-time”.

“Such a tool,” explains Spotify’s patent, “would allow artists to generate lead sheets more quickly and confidently by detecting and providing visual feedback as to whether any aspect of the work has a probability of being deemed plagiaristic.”

In short, Spotify reckons it can do better.

Spotify notes in its patent filing that “when executed manually, plagiarism detection is usually performed by experts and lawyers”.

One of the most well-known of those lawyers, is of course Richard Busch, who you’ll remember as the attorney that handled high profile infringement cases accusing Pharrell Williams and Robin Thicke of ripping off Marvin Gaye’s Got to Give It Up for their hit, Blurred Lines, and other infringement suits against the likes of Ed Sheeran and Juice WRLD.

Spotify’s patent application,  which you can read in full here, adds that “manual detection of music plagiarism requires substantial effort, skill and excellent memory, and is generally known to be impractical”.

Therefore, argues SPOT, “Software-assisted detection for text plagiarism on the other hand allows vast collections of documents to be compared to each other, making successful plagiarism detection much more likely”.

One of the tech’s credited inventors is a particularly interesting individual: AI expert François Pachet, who Spotify hired as Director of its Creator Technology Research Lab in 2017, poaching him from his as Director of the Sony Computer Science Laboratory in Paris.

Before joining Spotify, composer, scientist and AI music expert Pachet he oversaw Sony’s Flow Machines research project, which resulted in some of the first known contemporary pop songs composed with AI, including the Beatles-esque Daddy’s Car and the Duke Ellington-like The Ballad Of Mr Shadow.

By 2018, Pachet’s work culminated in the release of Hello World, which was described as ‘the first multi-artist commercial album created using Artificial Intelligence’.

When MBW interviewed Pachet in 2018, he explained that Artificial Intelligence should be embraced, not feared, by the music business.

Whether music rightsholders were convinced by that line of argument or not, his latest invention promises to add a new chapter to an already complex – and often fraught – relationship between Spotify and the songwriting community.Music Business Worldwide

Related Posts