Who Owns It? Who Made It? The Confusing Treatment of Works of Authorship Produced by Artificial Intelligence (Aug 2022)
Who Owns It? Who Made It? The Confusing Treatment of Works of Authorship Produced by Artificial Intelligence
By: Jeffrey Ke
There is a notorious thought experiment called the Infinite Monkey Theorem: if you give one billion typewriters to one billion monkeys and instruct them all to tap the keys rapidly and randomly, at least one of them with nearly 100% probability will type up the complete works of William Shakespeare at some point before the universe ends. In fact, every conceivable work of writing that will ever exist will be typed by these monkeys. One of the many questions this cloud of madly typing monkeys raises is, “Do the monkeys need to possess awareness of what they are doing to be considered the rightful authors of the works they produce?” Now consider if this cloud of monkeys were a six-foot tall supercomputer, whirring in some sterile white lab; this supercomputer, however, doesn’t spew out Shakespeare, Auden, or Dickens by pure chance—it writes its own novels, poetry, and news articles, with such legibility and convincing originality that you could swear it was written by a real human being, as opposed to mindless monkeys typing a mash of random words. But does the computer have awareness? Does it need it?
That six-foot tall supercomputer isn’t just a thought experiment anymore; it’s the current state of the art of Artificial Intelligence (AI) in the real world. The New York Times Magazine published an article documenting the vast distance AI machines called “neural nets” such as “GPT-3” have advanced in their journey to be able to mimic the English language with the subheading “OpenAI’s GPT-3 and other neural nets can now write original prose with mind-boggling fluency — a development that could have profound implications for the future.”11 Steven Johnson, A.I. Is Mastering Language. Should We Trust What It Says?, N.Y. TIMES (Apr. 15 2022), https://www.nytimes.com/2022/04/15/magazine/ai-language.html. GPT-3 has churned out online op-eds about the necessity of rights for robots,22 A robot wrote their entire article. Are you scared yet, human?, THE GUARDIAN (Sept. 11 2020), https://www.theguardian.com/commentisfree/2020/sep/08/robot-wrote-this-article-gpt-3. entire books of published poetry,33 Jukka Aalho, I Wrote a Book with GPT-3 AI in 24 Hours – And Got It Published, MEDIUM, (Jun. 11 2021), https://medium.com/swlh/i-wrote-a-book-with-gpt-3-ai-in-24-hours-and-got-it-published-93cf3c96f120. and scientific articles about itself.44 Almira Thunström, We Asked GPT-3 to Write an Academic Paper about Itself—Then We Tried to Get It Published, SCIENTIFIC AMERICAN, (Jun. 30 2022), https://www.scientificamerican.com/article/we-asked-gpt-3-to-write-an-academic-paper-about-itself-then-we-tried-to-get-it-published/. This thought experiment has become real, and we are just beginning to confront its implications on the institution of intellectual property rights. AI is emerging as a powerful decision-making and creative force, yet our copyright laws are still unprepared to deal with the question of an author being a machine.
The Current Treatment of AI-Produced Works of Authorship
The near-indistinguishability between many AI-produced works of art and literature and those of human authorship55 Sophie Nightingale & Hany Farid, AI-synthesized faces are indistinguishable from real faces and more trustworthy, NATIONAL LIBRARY OF MEDICINE, (Feb. 22 2022), https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8872790/. conveniently allows for the former to be judged by the current requirements of copyrightability—originality, fixation, and at least a touch of creativity. But if we were to stand in front of a watercolor fresco painted by an AI at an art exhibit, the programmer of the AI standing to the left of us, and the person who instructed the AI to paint to the right of us, we might ask ourselves, who’s the real author? If we were to apply the current framework of copyright authorship there would be three contenders for the title of author: the programmer of the AI, the end-user of the AI, and the AI itself. Though as we will see, handing authorship to any of these three parties under the current regime is problematic.
1. The Programmer as Author
We might first lift up the programmer as the rightful author of the works their machine generates. The programmer is certainly not undeserving; the years the programmer likely spent crafting the software program that would analyze works of art arguably amount to the largest contribution of discernible labor and creativity of all three candidates for authorship. It could further be said that the machine wouldn’t have been able to create if it were not for the ingenuity of the programmer; therefore, authorship of a machine generated work belongs with the machine’s creator.
However, the AI algorithm can generate the work without any input or direction from the programmer. In 2018 there was an AI-generated painting auctioned at Christie’s New York for $432,500 by three French students forming the art collective Obvious—without the permission or even awareness of the AI’s creator,66 Gabe Cohn, AI Art at Christie’s Sells for $432,500, N.Y. TIMES, (Oct. 25 2018), https://www.nytimes.com/2018/10/25/arts/design/ai-art-sold-christies.html. who had uploaded the AI code on the online code-sharing website Github with the intention of helping other AI-enthusiasts improve their own projects.77 James Vincent, How Three French Students Used Borrowed Code To Put The First AI Portrait In Christie’s, THE VERGE, (Oct. 23 2018), https://www.theverge.com/2018/10/23/18013190/ai-art-portrait-auction-christies-belamy-obvious-robbie-barrat-gans. That the programmer could be completely out of the picture in the creative process of the painting being conceived, drawn, and sold casts serious doubt on whether the programmer is entitled to copyright privileges, especially if there are many such paintings being made. Giving the programmer copyright privileges over all the works created by the independent AI they created would be granting monopolistic control over creativity and labor not their own. The faraway programmer has almost no hand in generating the creative works made by an AI; to long-distance mail them the profits of copyright protections would not incentivize them to produce more creative works—counter perhaps to the whole point of copyright law.
2. The End-User as Author
We might argue that the end-user is entitled to copyright for works generated by an AI for the same reason a photographer is entitled to the copyright for their photos taken by their camera. Despite the camera doing the physical work of fixation, i.e., capturing an image and pressing it onto the surface of digital memory, it is the end-user’s creative decisions that ultimately govern what is being captured by the camera. The iconic case of Burrow-Giles Lithographic Co. v. Sarony laid down the precedent for how copyright law should treat machine-assisted works by affirming that photographs are eligible for copyright protection; writing for a unanimous court, Justice Miller found that:
… by posing the [subject of the photo] in front of the camera, selecting and arranging the costume, draperies, and other various accessories in said photograph, arranging the subject so as to present graceful outlines, arranging and disposing the light and shade, suggesting and evoking the desired expression, and from such disposition, arrangement, or representation, made entirely by plaintiff, he produced the picture in suit.88 Burrow-Giles v. Sarony, 111 U.S. 53, 60 (1884).
Furthermore, handing copyright privileges to the end-user as opposed to the programmer would prevent monopolistic control over the large quantity of works an AI can make.
However, unlike a camera, artificial intelligence can produce original works of authorship without the end-user having to push a button. In fact, the end-user is almost entirely uninvolved in the creative process in most cases of AI-generated works, outside of merely clicking ‘START’; moments after a 2.7 magnitude earthquake hit Southern California on March 17, 2014, an algorithm called Quakebot extracted information from the United States Geological Survey reports and rapidly composed a brief news article documenting the event.99 Will Oremus, The First News Report on the L.A. Earthquake Was Written by a Robot, SLATE, (Mar. 17, 2014), https://slate.com/technology/2014/03/quakebot-los-angeles-times-robot-journalist-writes-article-on-la-earthquake.html. L.A. Times journalist Ken Schwencke merely walked over to his computer where the complete article was waiting for him and tapped a single key to publish it.1010 Id. Despite the appearance of the end-user initiating the process of generating the work, they are fundamentally no more involved than the programmer in the actual process of conception and fixation—making the categorical claim that end-users are entitled to copyright privileges to AI produced works also dubious.
3. The Machine as Author
By process of elimination, we arrive at the proposal that AI should receive the profits of copyright privileges. Rewarding AI as the actual force of creativity and fixation behind its works as we would human authors seems reasonable. Ben Kovach, writing for the Northwestern Journal of Technology and Intellectual Property, describes a type of AI called a “creativity machine” whose processes, while generating inventions instead of works of authorship, unequivocally demonstrate that it is the AI in charge of the creative process:
First, the creativity machine is exposed to a field of knowledge. Next, the machine is activated and begins to generate ideas. Finally, the machine flags ideas that are new and worthwhile… they are simply taught a general field of knowledge. They are not systems purposely built to perform a narrow task, they are simply taught and set on their way to generate whatever ideas they may generate.1111 Ben Kovach, Ostrich with Its Head in the Sand: The Law, Inventorship, & Artificial Intelligence, 19 Nw. J. Tech. & Intell. Prop. 137 (2021). However, a significant legal issue arises; machines and software have no legal personhood. According to the current copyright regime, there is in fact no entity to award copyright privileges to other than the programmer and the end-user. Furthermore, AI are simply not motivated by material wealth; handing a machine (or a computer software program) written certificate of its copyright would make it no more willing to produce or protect its creative works than if you had tried to stuff the certificate in its optical drive. Like an assembly line, an AI can crank out its creative works around the clock on command, needing no incentive of any kind. Giving AI copyright privileges would incentivize nobody to produce more creative works.
The Need for Change
At best, the current copyright regime’s treatment of AI-produced works of authorship leaves the question of ownership unanswered. At worst, it fractures the entire system, with the cracks only growing wider as AI becomes more prevalent in today’s economy.
1 Steven Johnson, A.I. Is Mastering Language. Should We Trust What It Says?, N.Y. TIMES (Apr. 15 2022), https://www.nytimes.com/2022/04/15/magazine/ai-language.html.
2 A robot wrote their entire article. Are you scared yet, human?, THE GUARDIAN (Sept. 11 2020), https://www.theguardian.com/commentisfree/2020/sep/08/robot-wrote-this-article-gpt-3.
3 Jukka Aalho, I Wrote a Book with GPT-3 AI in 24 Hours – And Got It Published, MEDIUM, (Jun. 11 2021), https://medium.com/swlh/i-wrote-a-book-with-gpt-3-ai-in-24-hours-and-got-it-published-93cf3c96f120.
4 Almira Thunström, We Asked GPT-3 to Write an Academic Paper about Itself—Then We Tried to Get It Published, SCIENTIFIC AMERICAN, (Jun. 30 2022), https://www.scientificamerican.com/article/we-asked-gpt-3-to-write-an-academic-paper-about-itself-then-we-tried-to-get-it-published/.
5 Sophie Nightingale & Hany Farid, AI-synthesized faces are indistinguishable from real faces and more trustworthy, NATIONAL LIBRARY OF MEDICINE, (Feb. 22 2022), https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8872790/.
6 Gabe Cohn, AI Art at Christie’s Sells for $432,500, N.Y. TIMES, (Oct. 25 2018), https://www.nytimes.com/2018/10/25/arts/design/ai-art-sold-christies.html.
7 James Vincent, How Three French Students Used Borrowed Code To Put The First AI Portrait In Christie’s, THE VERGE, (Oct. 23 2018), https://www.theverge.com/2018/10/23/18013190/ai-art-portrait-auction-christies-belamy-obvious-robbie-barrat-gans.
8 Burrow-Giles v. Sarony, 111 U.S. 53, 60 (1884).
9 Will Oremus, The First News Report on the L.A. Earthquake Was Written by a Robot, SLATE, (Mar. 17, 2014), https://slate.com/technology/2014/03/quakebot-los-angeles-times-robot-journalist-writes-article-on-la-earthquake.html.
11 Ben Kovach, Ostrich with Its Head in the Sand: The Law, Inventorship, & Artificial Intelligence, 19 Nw. J. Tech. & Intell. Prop. 137 (2021).