Coming from a computer science background, I find the intersection of technology and law interesting. Also having an interest in creative writing, I will explore the topic of AI-written books and copyright law.
Books written by AI
There are lots of books and stories written by AI. One example is Harry Potter and the Portrait of What Looked Like a Large Pile of Ash, which is a new chapter of the Harry Potter series that was created by AI trained on seven books of the Harry Potter series.[1] Another example is 1 the Road, which is a book written by AI based on sensory data inputted during a road trip.[2] This second example is a little bit different from the first. In the first example, AI produced the book after being “fed” other books, but the AI in the second example produced its book based on the physical sensory input that it got exposed to through a camera, GPS, microphone, and clock.[3] A third example is Amor Cringe, which is a novel co-written by an AI and a human author.[4] These three examples show some of the diverse ways AI can write novels with current technology.
Should an AI’s work qualify for copyright?
We know that regular novels written by humans can be copyrighted. A novel written by multiple authors can also be copyrighted, and each author would have copyright over that same novel. As well, we know that AI can write actual books with sentences, chapters, and a (sometimes) compelling plot. Although the quality of the books written by AI may be lower compared to some of the books written by humans, low-quality books by humans are also subject to copyright, so the quality isn’t an issue when comparing books by humans and books by AI. This means that the only difference between a book written by a human and a book written by AI in a simple sense is whether the author of the book is a human or an AI, which, in itself, encompasses many differences.
Section 5(1) of the Copyright Act states that any original literary, dramatic, musical, or artistic work can be copyrighted.[5] The point of analysis is the “originality” requirement. As per CCH Canadian Ltd. v. Law Society of Upper Canada,[6] originality requires the exercise of skill and judgement. So, can AI exercise skill and judgement?
Some would argue that a “machine” is incapable of exercising skill and judgement. It simply operates according to the input data it is given and performs actions that it is coded to do. This can be thought of as purely mechanical, so it wouldn’t be considered an intellectual effort akin to what humans do when producing a literary work.
On the other hand, I’d like to pose a question: Don’t humans and AI think in a similar way? All the experiences and knowledge that people gather throughout their lives is like Big Data.[7] We use this gigantic database to form opinions, evaluate, and compare. Further, we consistently improve ourselves based on this data, our own actions, and outcomes of our own actions. We don’t exactly know how we turn this data into all the complex things we do, but it just happens. AI acts in a similar way. It is fed data and uses that to make calculations and produce outcomes. With machine learning, AI can constantly improve itself based on the input data, new data, and data produced by its own actions.[8] Again, we don’t really know exactly how AI does all this, but then again, it’s the same for humans as well. So, moral considerations aside, aren’t there some striking similarities between the ways in which intelligence works for humans and AI?
Another argument against AI’s work being able to be copyrighted might be that AI produces its work based on similar works it is fed. So, AI’s work can be seen as more like a mash of humans’ works it is inputted and thus lacking originality. However, isn’t this sort of like how humans produce work? Like mentioned in the paragraph above, humans have a repository of experiences, including novels read during a person’s entire lifetime. The outcome of our writing is based on this cumulative knowledge that we gained, similar to how an AI’s output is based on data that is inputted. Nevertheless, one big difference would be that humans have creativity – we can get ideas that are different from anything they have experienced in the past, or we can work with ideas in a creative way. Maybe AI can get better at generating creative ideas and solutions, but in my opinion, creativity is more of a human element as of now. Another difference is emotions; humans have emotions, but current technology doesn’t allow for AI to truly have emotions. AI can generate work that triggers human emotions, but AI itself does not have feelings. Therefore, work produced by AI is produced in a different way than work produced by a person.
Considering all the discussion above regarding whether a work by AI should have copyrights, I think the answer as of now is that works by AI cannot have copyright in Canada. Courts have implied that copyright has a human aspect to it,[9] and more research is needed to establish that work by AI can be original and that AI can exercise skill and judgement. However, I expect that works by AI would be able to be copyrighted in Canada in the future, whether it be changes in law that would allow for it or development in technology and novel research findings.
If works by AI can be copyrighted, who should own the copyright?
If AI produces work having only been inputted work by one author, like in the case of Harry Potter and the Portrait of What Looked Like a Large Pile of Ash, the number of candidates of people who might be entitled to copyright is smaller. In the example above, the author of the Harry Potter series would be one candidate, the ones who made the AI would be another, and the person who used the AI to produce the work might be another candidate. The problem is that it’s unclear who the “author” is.
The issue gets even more complex when the input data of the AI is from a large database. With a large database, countless people’s work would be inputted into the AI. For an AI that writes books, maybe all the books in the world that are available online can be inputted into the AI. On top of that, not only books might be inputted, but anything else that is written (e.g. tweets, social media posts, blogs, academic writing, etc.) may also be in the database. In this case, the candidates for the copyright owners of work by an AI fed all this data would be: (1) everyone whose work/writing is in the database; (2) the ones who made the AI; and (3) the ones who used the AI to produce that work. I think that by this point, the work by AI should just be in the public domain because of the sheer number of people who contributed to the work.
On the other hand, maybe the authors of the works inputted into the AI shouldn’t have copyright for the work generated by AI in the first place because the work produced by AI is a totally different work than the work inputted. Although an AI might be fed 100 books, the novel that the AI produces would have different characters and storylines as the 100 books it was fed to produce that work.
In summary, the question of who should own the copyright of an AI-generated work is complex. My opinion is that as technology develops, AI would likely be inputted large datasets with countless contributors instead of a small dataset containing work by only a few people. Since it’s impractical and inefficient to attribute copyright to countless people for every single work that countless AI produces, it would be better to attribute work by AI to the public domain. However, this all depends on the specific situation, and there may be cases where it’s better to attribute copyright to people for AI’s work.
Should AI be able to have its own copyright?
Today, even a corporation is a “person” by law.[10] Therefore, with enough sophistication and development in technology, I think it would be possible for AI to have its own copyright sometime in the future. However, that is not the case right now. Before deciding whether an AI should have its own copyright or not, there should be more advancements in technology that enable us to answer the questions of whether work by AI can even be copyrighted, and if it can be, who should own the copyright.
*Professor Festinger, I’d like to submit this as my final paper for the course. Thank you.*
[1] See “A hilarious new Harry Potter chapter was written by a predictive keyboard — and it’s perfect” (12 December 2017), online: Mashable <https://mashable.com/article/harry-potter-predictive-chapter>.
[2] See “AI-Written Books: Can Artificial Intelligence Write a Novel?” (10 October 2022), online: BeInCrypto < https://beincrypto.com/ai-written-books-can-artificial-intelligence-write-novel/>.
[3] See “1 THE ROAD BY AN ARTIFICIAL NEURAL NETWORK” online: JBE Books < https://www.jbe-books.com/products/1-the-road-by-an-artificial-neural>.
[4] See “A Novelist and an AI Cowrote Your Next Cringe-Read” (24 May 2022), online: WIRED < https://www.wired.com/story/k-allado-mcdowell-gpt-3-amor-cringe/>.
[5] Copyright Act, R.S.C. 1985, c. C-42.
[6] CCH Canadian Ltd. v. Law Society of Upper Canada, 2004 SCC 13.
[7] See “What Is Big Data? Definition, How It Works, and Uses” (29 November 2022), online: Investopedia <https://www.investopedia.com/terms/b/big-data.asp>.
[8] See “Machine Learning” (15 July 2020), online: IBM <https://www.ibm.com/cloud/learn/machine-learning>.
[9] Supra note 6.
[10] Business Organizations Lectures at the Peter A. Allard School of Law (2022).
Communications Law