Ars Ex Machina: Copyrights and Artificial Intelligence

October 4, 2019

By David Posteraro and Alex Welsh

intellectual propertyWe are at or near an inflection point where the computer programs we use to create are no longer tools or instruments of the author, but arguably authors themselves. So, who owns the copyright on computer-generated works? This is not a mere curiosity. Artificial intelligence “will have a profound impact on the world, transforming the way we live and work” and software is “eating the world,” as the 111 billion lines of code generated every year run more and more of our critical systems. As that software approaches the authorship tipping point, it is important to understand the state of the art and the law that governs ownership of its creations.

State of the Art

True human-level general intelligence is not currently replicable with programming. Indeed, that reality has been “20 years away for 60 years” and experts remain “very skeptical” that it will arrive in the next few decades. One need look no further than the struggles of self-driving cars to see that computer-based human intelligence is indeed still a “distant dream.”

So what can artificial intelligence do? While current intelligent systems fall down in areas outside of their specialized programming, they excel at narrowly-tailored tasks. In those well-defined roles, artificial intelligence is rapidly becoming integrated into our daily lives – in web searches, navigation programs, music recommendations, mobile banking, commercial flight and many other areas.

But are these programs more author or tool? An illustrative example is the recent album created by the band Yacht with the help of Google’s MusicVAE. The A.I.-assisted creative process took three years of intensive work by the band. First, Yacht had to train the algorithm, and convert its back catalog to a form the program could process by notating all 82 songs in MIDI format. Yacht then used the program as a tool to generate countless components of songs – riffs, lyrics, titles, etc. – based on the inputs. The band then curated the usable ideas and sent the rest back to “the multidimensional mathematical obscurity from which it emerged.”

It might seem, then, that the A.I. has no real authorship in the album. Not so, according to Yacht. From their perspective, the A.I. developed riffs that they don’t know that they could have written themselves, noting that “It took a risk maybe we aren’t willing to take when we’re writing a pop song.” As such, Yacht characterize the A.I. as a “constant collaborator.” How, then, does our current legal framework square with the capabilities of modern computing?

The Law                                                                                        

In the United States, the issue of non-human authorship appears, at first, to be resolved by a bright-line rule. The Compendium of U.S. Copyright Office Practices explains that “because copyright law is limited to original intellectual conceptions of the author, the Office will refuse to register a claim if it determines that a human being did not create the work.” That position is quite clear. However, the Register further explains that “the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author.” This policy leaves plenty of room, however, for work of the type created by Yacht, which seems to involve substantial creative input from human authors.

Yacht selected the inputs used to train the algorithm (collecting over 2 million words of lyrics), selected usable content from among the program’s prolific output (hundreds of riffs took only a few minutes to generate), and performed the works based on the “musical instructions” created using the A.I. Under no reasonable interpretation can that process be considered merely mechanical or lacking the creative input of a human author. Yet, to some extent, Yacht went out of its way to avoid altering the A.I.’s output. The band imposed strict rules for working with the algorithm, including that they did not improvise or harmonize and that only subtraction was allowed. So the A.I.’s contribution can’t be denied, either.

This distinction may seem unimportant, but it is one that will need to be fully explored as artificial intelligence becomes more sophisticated, creative and pervasive. We will be forced to wrangle with the notion that authorship is no longer a purely human endeavor. Other jurisdictions have taken proactive steps to acknowledge the coming boom in computer-authored works. For example, the United Kingdom recognizes computer-generated works with no human author and grants copyright protection to “the person by whom the arrangements necessary for the creation of the work are undertaken.” The European Parliament considered, but ultimately rejected, granting computer programs the status of “electronic personhood” in the copyright context.

For now, A.I. requires human involvement in the creative process, making legal acknowledgement of non-human authorship a non-issue. But we are approaching a period of independent creation by intelligent systems, and our laws may not be properly calibrated to adequately protect such works. In other words, it’s about time we realized the piano can write the concerto.

For more information regarding this article or similar topics, visit KJK’s Intellectual Property practice group page, or contact David Posteraro at drp@kjk.com or 216.736.7218, or Alex Welsh at amw@kjk.com or 216.736.7263.


KJK publications are intended for general information purposes only and should not be construed as legal advice on any specific facts or circumstances. All articles published by KJK state the personal views of the authors. This publication may not be quoted or referred without our prior written consent. To request reprint permission for any of our publications, please use the “Contact Us” form located on this website. The mailing of our publications is not intended to create, and receipt of them does not constitute, an attorney-client relationship. The views set forth therein are the personal views of the author and do not necessarily reflect those of KJK.