SPECIAL OCCASIONS
S/O STOCK
Web, Publication, Video
2024
"SPECIAL OCCASIONS" captures the unique imperfection of technology and how they shape our perception of reality. The most cliché scene of a family of four dining together is depicted as a special occasion in this idealized reality generated by AI. In a world filled with surreal images, this work focuses on creating the most ordinary and realistic images as represented by technology and then tweaking them through the perspective of AI. This approach allows us to feel a range of emotions, from warm memories to alienation, as technology tries to accurately depict human society.
https://doh2ht.github.io/SO-STOCK/
Remove Nipples
Publication
2024
How does AI view and regulate women's bodies? I've noticed that while AI claims to protect us from harmful content, it often excessively restricts images of women's bodies. Ironically, it also generates more pornographic content through its algorithms. This reflects a continuation of the male gaze, and 'Remove Nipples' shows how society still imposes control over women's bodies through AI censorship.
Deceptive Magazine
Publication
2023
Editorial Design for the article 'ChatGPT is a blurry jpeg of the web (Ted Chiang)' in the magazine 'Deceptive Magazine'
labeled noise
Installation, Web, Publication
2019
The way artificial intelligence reads images is not yet complete, and it creates a strange sense of confusion because there is a difference from the way humans think. “labeled noise” is a visual research that defines noise, an imperfect thinking process of artificial intelligence, and archiving the the result of image recognition. The new search term classifies the image as noise while also becoming the new label for the image. As the same image is interpreted as a completely different label, we question which one should be labeled on the image. Depending on the number of operations of the image iteration, the message is distorted, regardless of its intention. When the search term is derived from the apple to the apple again and delivered without distortion of meaning, no noise exists, and a new label is generated only when noise exists.