Matthew Vo, 12th Grade

The past decade has experienced dramatic growth of interest in Artificial Intelligence (AI) programs, with particular attention focused on “creative AI”, capable of producing visual artwork, writing, or music at a level of quality rivaling that of works created by human artists and authors. However, such developments have also raised pressing legal and ethical concerns, as well as philosophical questions on the very underpinnings of “human” artistry which these machines now supposedly emulate.

Given the capabilities of “creative AI”, much discussion has been raised debating the extent to which such programs can be considered “creative”. All modern AI consists of neural networks, computational systems loosely based on neurons in the human brain, which “learn” by analyzing large quantities of data for patterns. In many art-creating AI programs, this basic function has been enhanced through Generative Adversarial Networks (GANs), which improves the realism of the AI’s output by using two neural networks – one to create output, and the other to detect whether the output appears “false” (1).

Considering the above processes, Marian Mazzone, Associate Professor of Contemporary Art at College of Charleston, SC, argued in a 2019 essay that the machine learning approaches utilized by GAN-based AI display a striking parallel to human creative thinking, pointing out that the way the two neural networks worked to balance the stylistic consistency and novelty of the outputted creations greatly resembled the manner in which human artists use visuals to synthesize original creative output (2). Conversely, others affirm that AI programs currently lack the ability to possess creativity, particularly because they lack sentience and other innate qualities of human artists. Jon McCormack, head of the SensiLab technology research group at Monash University, Australia, argues that GANs merely mimic patterns in the training data and being preprogrammed, do not possess free will, motivation, and spontaneous life experiences from which to create art. (3).

It is important to note that the question of creativity itself is highly subjective, not least because the concepts of “artwork”, “artists”, and the process of creating art are all subject to individual preconceptions and prior experiences, thus complicating a clear-cut answer to the question of machine creativity (4). Furthermore, the popular connotation suggested by the term “artificial intelligence” also plays a role, as non-experts in the field tend to think of AI, misleadingly, as functioning substantially similar to human intelligence (3). Thus, the idea of “creativity” as it pertains to AI will remain a subject of intense debate for some time.

Far more significant are the legal ramifications of AI-created art, posing new challenges for intellectual property protections such as copyright law. Broadly speaking, the fundamental purposes of copyright are to provide economic incentives for continued creative production that benefits wider society, protect authors’ rights to “the fruits of their own labor”, and defend aspects of authors’ identities that are embodied in their creations (5). Crucially, these principles naturally imply the creative author in question is human, prohibiting non-humans such as animals, mechanisms, and the environment from holding copyright to creations in most legal systems (6). In the context of AI-created art, this requirement is highly consequential, as AI-created works are seldom protected by copyright in the U.S., providing a potential disincentive for artists and software developers to continue pursuing innovation in the field (1).

To this end, several opposing perspectives for approaching intellectual property protection of AI-created works have been proposed, one being the granting of copyright to the AI developers. Such an approach would be reasonable since programmers arguably determine the AI’s behavior and in turn, any outputs it creates (7). However, opposition has focused on the fact that programmers merely create the framework for which the AI operates, with little to no control over the subsequent ways the program “learns” independently through its algorithms. Alternatively, it has been proposed that credit be given to users of the AI, since they personally controlled the output to their own specifications (6). However, that assumption may be an overgeneralization, as the amount of direct involvement and interest the user has in the AI’s creation process may vary widely across individuals or circumstances (7). Some arguments have gone further, proposing that AI programs themselves should hold copyrights to their creations. This makes sense, since the AI actually “created” the work (6) and can be seen as possessing a form of “creativity” analogous to its human counterpart (2). Nevertheless, granting copyright ownership to a nonliving algorithm would be an unprecedented leap, since it would require the currently-universal “human author” dogma be overturned and necessitate a reevaluation of the purposes of copyright protection.

In addition to legal ambiguity, the sheer effectiveness at which AI-created works can now match their human-created counterparts are fueling concerns that AI may increasingly supersede human artists. Indeed, as pointed out by Dr. Fabio Morreale of the University of Auckland’s School of Music, future commercial adoption of AI-produced music by streaming and distribution platforms such as Spotify would risk exacerbating the struggles of a vast majority of music artists to gain online viewership (8), with dire consequences for those pursuing an inherently unpredictable, competitive career path. Furthermore, “creative AI” may potentially worsen the spread of misinformation, as noted by Allen Institute for AI CEO Oren Etzioni regarding DALL-E, a recently developed AI program which produces photorealistic images based on user-inputted phrases (9). In fact, a recent study found that human participants were not only unable to distinguish AI-generated images of human faces from real ones, but also tended to rank the AI-created images as “more trustworthy” (10).

Ultimately, it is evident that AI-created art, typical of transformative new innovations, poses new challenges while altering long-held precedents, particularly within the legal, artistic, and philosophical spheres. Although the technology remains largely experimental and is still actively being developed, such “creative AI” should be expected to become increasingly commercialized and widely available. As that occurs, the issues discussed above, many of which are currently theoretical, will likely become pertinent, thus only increasing the relevance of these debates in the coming years.

Citations:
[1] N.I. Brown, Artificial authors: A case for copyright in computer-generated works. Science and Technology Law Review 20(1), (2019). doi: 10.7916/stlr.v20i1.4766
[2] M. Mazzone & A. Elgammal, Art, creativity, and the potential of artificial intelligence. Arts 8(1), 26 (2019). doi: 10.3390/arts8010026
[3] J. McCormack, T. Gifford, & P. Hutchings, Autonomy, authenticity, authorship and intention in computer-generated art. Computational Intelligence in Music, Sound, Art and Design, 35-50 (2019). doi: 10.1007/978-3-030-16667-0_3
[4] M. Coeckelbergh, Can machines create art? Philosophy & Technology 30, 285-303 (2017). doi: 10.1007/s13347-016-0231-5
[5] M.E. Kaminsky, Authorship, disrupted: AI authors in copyright and First Amendment law. UC Davis Law Review 51, 589 (2017).
[6] T. He, The sentimental fools and the fictitious authors: Rethinking the copyright issues of AI-generated contents in China. Asia Pacific Law Review 27, 218-238 (2019). doi: 10.1080/10192557.2019.1703520
[7] R. Matulionyte & J. Lee, Copyright in AI-generated works: Lessons from recent developments in patent law. SCRIPTed 19(1), (2022). doi: 10.2966/scrip.190122.5
[8] F. Morreale, Where does the buck stop? Ethical and political issues with AI in music creation. Transactions of the International Society for Music Information Retrieval 4(1), 105-113 (2021). doi: 10.5334/tismir.86
[9] C. Metz, Meet DALL-E, the A.I. that draws anything at your command. The New York Times, (2022).
[10] S.J. Nightingale & H. Farid, AI-synthesized faces are indistinguishable from real faces and more trustworthy. PNAS 119(8), (2022). doi: 10.1073/pnas.2120481119