Hsuan-Hwa Chu, MA Visual Effects
I. Introduction
This is a blog for “Stylisation Without Substance: A Critical Examination of Tool”. It covers my personal thoughts, reflections, and writing process for this article. This blog does not adopt an academic perspective, so the language used may appear more casual.
II. Background
I must acknowledge that my initial objective was to compose an article that would explore the intersection of Japanese otaku culture and social interactions with contemporary video production. After discussing this subject with Nigel, I received feedback at the time that suggested I should concentrate on a more specific subject and that the audience would benefit more from it.
Following that, I began to contemplate the “tools” that authentically influence our current creative output. The immediate response was the implementation of AI technologies. KlingAI and DoubaoAI are examples of generative AI that I have personally employed. At first, I was both surprised and slightly apprehensive. I am even more awed by the development of video AI. The images’ flawlessness and beauty, as well as their ease of creation, seemed peculiar. It prompted me to consider my role in this process, given the exceptional quality of creation.
III. Research Methods
I approached this critical report not only as a VFX student but also as a systems thinker, given my academic background in the social sciences. I was able to investigate generative AI beyond its functional outputs as a result of my experience in analysing social structures and institutional authority. I investigated the politics of authorship, value systems, and embedded hierarchies.
This dual perspective significantly influenced my research process. Not only did platforms such as Google Scholar, ScienceDirect and JSTOR become indispensable for obtaining technical insights, but they also proved essential for identifying high-impact critiques. To comprehend the epistemic and cultural repercussions of algorithmic tools, authors such as Burrell (2016) and Birhane (2021) were indispensable. I concentrated on the ways in which AI reshapes creative labour, cultural legitimacy, and interpretive authority, rather than merely describing how it operates. Technology and society intersect with these concerns.
III. Research Process
As I researched, I found that many AI systems are educated on massive internet databases. That sounds neutral but isn’t. The datasets’ cultural biases are incorporated into the AI’s outputs. Sources suggest Black women are oversexualized and Asian women are fetishised. This struck me as a social issue, not simply a tech one. So I changed my essay’s focus. I started looking at what’s lost in AI-generated art instead of just admiring it. Human context. Cultural nuances. Psychological intent. AI may create stylised visuals, but are they meaningful? Writing the essay was hard. I used these technologies myself, so I had to balance criticism and fairness. I didn’t want to blame technology. Not “AI bad, human good”. What kind of civilisation are we creating if we use tools that aestheticise everything but understand nothing?
I was struck by an article on how generative models “extract” creativity from people. Then it clicked. These gadgets shape us, not just help. They affect our taste, workflow, and judgement of good and poor. They often do this covertly. I’m glad I switched topics. This writing feels urgent and real. It also shows my passion for why humans produce things and what it means when a machine answers that question. Further investigation revealed a Birhane et al. (2021) study that astonished me. The researchers uncovered harmful stereotypes present in large image datasets, such as LAION, which are used to train technologies like Stable Diffusion, particularly regarding the sexualisation of Black and Asian women. This went beyond technical bias. This was about automated cultural violence. Birhane et al. (2021) That transformed my view of “creative tools”. They recognised they shape what pictures exist, not simply help. Young’s (1995) “Five Faces of Oppression” is a theory I learnt in my undergraduate studies, and it is very appropriate to use here.
I thought, certainly, these visuals are “stylised”—but what’s behind them? They have what cultural memory? What human emotion or struggle? Burrell (2016) made an excellent point about how algorithmic systems might appear “opaque” or impartial to users but are deeply political in how they filter, classify, and privilege data. That was especially important in generative art, because the training set filters creativity. The further I explored, the more I discovered this wasn’t simply about tools. The concepts of labour, power, and cultural authorship all play a significant role.
Additionally, among all the articles I evaluated, Essien et al. (2024) was the closest to my topic. This paper, unlike others on AI in creative industries, focuses on Hollywood VFX, making it relevant to my MA in Visual Effects. AI-assisted workflows and human artistic labour interact in complex ways, with the authors claiming that creative intent and interpretation decisions still depend on human expertise. Their emphasis on interpretive instability—how artists must make meaning when AI-generated visual effects lack clear connections to real-world events or narratives—moved me most. My thesis on algorithmic opacity and the requirement for human authorship in framing and defending the work was directly influenced by this thought. This paper supported and refined my argument. It connected theoretical AI critiques to my daily digital production experiences.
IV. Reflection
Overall, I believe that I may have underestimated the amount of time that this type of critical writing actually requires. At the outset, I dedicated a significant amount of time to contemplation and reading; however, I neglected to allocate sufficient time to refine and refine the fundamental argument. Certain sections, particularly those in the middle, could have benefited from more robust logic or more effective transitions. It was too late for me to recognise that simply possessing favourable references is insufficient; it is also necessary to establish a clear connection between them and your argument.
One example: I acknowledged the bias in AI datasets; however, I did not provide a comprehensive explanation of how this bias directly influences the “stylised” content that these tools generate. The correlation was present in my mind, but it may not have been as prominent on the page. If I had had additional time, I would have strengthened those connections and potentially incorporated more personal or visual examples.
Nevertheless, I remain pleased with my decision to select this subject. It encouraged me to consider the systems that underpin creativity today, rather than limiting myself to superficial aesthetics. And despite the imperfections in the writing, the process facilitated my increased awareness of the manner in which I wish to interact with these tools—not merely as a user but as an individual who continues to prioritise intention and significance in the creative process. I will undoubtedly prioritise time management in the future, and I may even construct the structure in advance rather than following inspiration halfway through.
V. Presentation Video
Leave a Reply