Is AI a threat to human creativity?
Institute for Ethics in AI Bernard Williams Essay Prize | Sep 30, 2024Introduction
I believe AI poses various threats to human creativity, but is all-in-all not a threat in itself. I believe human perception is the central differentiator between AI and human creativity; AI cannot fully replace human creativity unless it develops the ability to perceive, think and feel autonomously, which, given the nature of AI, is unlikely in the foreseeable future. I argue AI poses two distinct kinds of threats to human creativity − potential to outperform, and potential to disrupt.
Potential to outperform
The product-focused view of creativity defines creativity or creative ability by the qualities possessed by tangible outcomes of a creative process, e.g. artworks, inventions or solutions. Popular philosophical view deems an object or idea creative ‘if and only if it is novel and valuable’ [1] [2].
Artistic creativity
In art (including visual art, music, literature etc.), creative novelty can be understood as how much it deviates from what has already been made, characterized by a unique perspective, style and/or composition. It is difficult to imagine how AI may deviate from what is already made when AI is trained on a database of existing works; perhaps through a unique combination of existing works and styles can it create novel artwork. After all, there is a distinct look and feel associated with AI art unlike anything humans have previously created, albeit being described as ‘mediocre’, ‘uncanny, smooth, and generally off’ [3]. That said, novelty alone does not guarantee creative value; artwork should also possess high artistic value to be deemed 'creative'.
Creative value in art is highly multifaceted: artistic value is understood by a tapestry of measures, including (but not limited to) aesthetic, emotional, personal and cultural value. Aesthetic value can be understood as the extent to which an artwork is 'beautiful', or pleasing to the senses. This can be subjective, but established artistic principles often guide and influence our perceptions of beauty and harmony. AI art generators are considered to currently be in a rather unsophisticated form, generally associated with artificiality, logical inconsistencies and an overall lack of harmony and cohesion. However, neural network image generators are exponentially improving in the sophistication and quality of image generation, generating images that are closer to being indistinguishable from that created by human artists. For instance, a quiz from an art blog [4] testing quiz-takers' ability to tell human art apart from AI managed to trick over 50% of quiz-takers in one of the questions (see below; the left artwork is human-made). Some artworks have even received institutional recognition, from the first AI-generated artwork Edmond de Belamy valued at US$432,500 [5] to the grand-prize-winning Théâtre D'opéra Spatial [6].
Whether AI artworks, including those institutionally-recognized, are truly deserving of artistic merit remains contentious; I argue that the nature of 'high-value' AI art is of a different nature than human art in that it lacks artistic sense from genuine perception. This human artistic sense may sometimes be replicated by chance, or by quality and refinement of the prompt, but such a replication is not indicative of true artistry.
Additionally, aesthetic value does not accommodate all artistic values. The emotional value of art is understood as the emotions conveyed through the artwork; while personal value focuses on the personal significance carried by the artwork - a sense of personal authenticity, the unique circumstances, experiences, relationships, emotions, etc. shaping the artwork. Clearly, AI cannot convey emotion or possess agency by virtue of being a machine.Cultural value focuses on the influence an artwork has on cultural discourse, overarching trends, etc. While AI art may have this effect, it is not in the inherent value of the art produced but rather the discourse around the subject matter of AI art that creates such cultural value.
Ideation creativity
The creation of ideas, or ideation, is an inherently different process than the creation of art. Novelty and value are defined differently: ideation novelty is understood as a unique perspective on, and/or synthesis of existing information, while the value of an idea is attributed to its ability to enhance the quality of human life. Similarly, Guilford’s model of creativity [7] delineates the creative process as an interplay between divergent thinking (ability to make unusual (reasonable) associations) and convergent thinking (ability to identify the most useful and relevant ideas, and put information together in a way that creates added value).
Divergent thinking is typically and intuitively thought of as a domain where humans reign supreme, but studies have shown AI outperforming humans in a divergent thinking task [7]. However, the nature of AI caps its divergent thinking ability at the quality of its inputs and the parameters by which it trains; additionally, its inability to genuinely experience the problems it sets out to solve or contemplate is what bars the novelty of AI ideas from reaching the upper limit of human idea-generation novelty. Human problem-solving and inspiration stems from human observation of the world, from the development of various modes of transportation to the smartphone to a kitchen gadget.
Convergent thinking is a domain where AI is meant to excel in, due to its efficiency, ability to retrieve a wide database of information, its training in recognition of patterns on labeled datasets, and its precision and freedom from human bias. However, it is not a faultless replacement: AI is unable to engage in nuances or offer a truly fresh perspective. I put this exact essay prompt through Perplexity to compare its ability to consolidate ideas with my own [8]: while Perplexity offers a clear stance, discussing key debates and points of interest, and offers a concise and balanced point of view, it is merely a summary of existing debates. My essay, while perhaps less straight-forward, offers a more nuanced thesis, abstract conceptualizations with the separation of threat into outperformance and disruption, or the inclusion of anecdotes and examples, attempting to add perspective and get to the root of the issue.
All in all, AI excels in convergent thinking as it excels in synthesizing information and recognizing patterns, but is unlikely to exceed the upper limits of divergent thinking, and lacks the perspective to produce truly boundary-pushing ideas whether by convergent or divergent thinking.
Threat as disruption
Job displacement
Firstly, AI disrupts human creators by displacing creative jobs. It is a truism in business that companies aim to maximize profit and minimize cost: with the introduction of AI, companies are less likely to hire freelance designers, copywriters, etc. to create professional designs or writing, especially startups or SMEs who are less able to afford such services. Instead, such companies turn to AI platforms as a replacement for professional human creatives, especially as most platforms providing AI services are free of charge (e.g. Canva, Adobe Sensei or Copy.AI). A surplus of professional creatives are therefore rendered unemployed due to an influx of AI replacements.
In the commercial world, art is not judged on novelty and value [3] as aforementioned in the art world, but rather viewed as a saleable commodity; its value is defined by its ability to bring profit to the company. AI researchers say AI was intended to 'democratize image-generation for a bunch of people who wouldn't necessarily classify themselves as artists', or rather, provide a shortcut to creativity and significantly lower barriers to entry. While seemingly well-intentioned, this commodifies and devalues art in itself.
Authorship
Secondly, AI also disrupts human creators by complicating authorship/intellectual property (IP) matters. AI is trained on a database of artworks/music/written works, some of which were copyrighted and all of which were originally created by a human artist/creator - many of whom were not asked for permission to use their work in the training of AI. The current state of legal protections surrounding AI and authorship is relatively weak: people cannot try to copyright AI-generated work due to lack of human IP ownership, unless significantly modified. Given this, any application for copyright protection requires disclosure of any AI usage - failure to do so simply results in denial of copyright. In any other cases, disclosure of AI usage is not explicitly legally required, and there exist little to no consequences of failing to do so. It is also generally legal for AI companies to train on copyrighted works.
Conclusion
To conclude, the products produced by AI can sometimes exceed the novelty and value of that created by humans, but the processes by which they are created lack a genuine perception of the outside world, nor are they created by autonomous agents capable of conscious decision-making.
AI is also more likely to outperform humans in the production of creative ideas than the production of art, as artistic value is more contingent on process and/or agency factors, which AI cannot emulate, than the value of ideas. If we understand art as a celebration of the lived human experience through an aesthetic medium, and ideas as solutions to enhance quality of life or provoke thought, it is clear as to why AI excels in producing the latter.
In light of the above findings, humans should leverage their unique perception to push boundaries of human creativity and remain ahead of the curve; recognize AI's strengths and weaknesses, leveraging the precision and consolidation strengths of AI ideation to solidify foundation of ideas; upskill by learning the art of maximizing value gained from AI, and when mastered, slowly integrate the artful, proper use of AI into education, and tighten regulations surrounding the ethical use of AI to protect artists' rights.
Bibliography
[1] sciencedirect.com/science/article/pii/S0039368123000833
[2] plato.stanford.edu/entries/creativity/
[3] https://www.vox.com/culture/351041/ai-art-chatgpt-dall-e-sora-suno-human-creativity
[4] https://foundmyself.com/blog/ai-art-quiz
[5] https://www.christies.com/en/stories/a-collaboration-between-two-artists-one-human-one-a-mac hine-0cd01f4e232f4279a525a446d60d4cd1
[6] https://www.nytimes.com/2022/09/02/technology/ai-artificial-intelligence-artists.html
[7] https://www.nature.com/articles/s41598-023-40858-3
[8] https://www.perplexity.ai/search/write-a-1500-word-essay-respon-HG9EbuNFRzq.ydshUFzwfA
