Unreal Pareidolia -laser caustics-

Exhibited at studio doué (2023)


 There is a phenomenon called pareidolia, which is the tendency for the mind to recall familiar patterns when it sees something, even though they do not exist in the original image. This phenomenon is the act of imagining a certain object, for example, “the clouds look like ice cream and look delicious” or “the shade of the tree looks like a person and is scary. In Japan, it has long been used as a way of “seeing” and has permeated various cultures, including the tea ceremony and Japanese gardens. Pareidolia is also used in Rorschach tests to assess the psychological state of patients by asking them what they see as ink stains. Richard Gregory conducted a drawing test (reversing Rorschach test) that induced pareidolia and showed a correlation between the subject’s interest in art and the originality of the pareidolia. In other words, there is a relationship between having an original point of view and art.
 When people make a judgment, they search for visual similarities in their own memories. Thus, it can be said that the process of perception depends on our memories and experiences, as well as on our ability to imagine them visually. Therefore, we thought that we could expand the viewer’s perspective by creating an external device that is not dependent on these factors.
 In this work, the image and its caption are generated by using img2img, which does not dare to write prompts in the generation AI, for the pattern of reflected light from a laser beam hitting an everyday object such as OPP tape. The image is generated by Stable Diffusion and the caption is generated by BLIP-2. The user first observes the laser pattern. The caption is then displayed on the underside of the laser pattern, and by reading it, the user can confirm what the AI’s interpretation of the laser pattern is and think about which part of the laser pattern it actually is. The images generated by the AI are then projected onto the laser pattern by a video projector and superimposed on the laser pattern, allowing the viewer to see the correspondence between them. After the experience is complete, the laser is moved to a different position, and the above experience is repeated many times with different patterns.
 In today’s society, various systems are becoming more sophisticated and streamlined, making our lives more convenient. In particular, AI-based systems, such as translation systems, facial recognition systems, and generative AI, have made remarkable progress in recent years. While these AI systems push forward optimization, people depend on AI for their abilities, and all output by people in the world is expected to become homogenized. In such a situation, this work functions as an entity that influences not the output of people themselves, but rather the cognitive aspects that occur in each of them, and draws out their imagination. This work shows how to build a relationship with AI that will ensure the diversity of people.

 本作品はレーザーがOPPテープなどの日用品に当たった反射光のパターンに対して,生成AIにおけるプロンプトを敢えて書かないimg2imgを使用した画像の見立てが行われ,画像とそのキャプションが生成されます.画像生成はStable Diffusion,キャプション生成はBLIP-2により生成されます.体験者はまずレーザーのパターンを観察します.その後AI側の見立てによるキャプションがレーザーパターンの下側に表示され,それを読むことで,AIの解釈では何に見立てられたのかを確認し,実際それがレーザーパターンのどの部分にあたるのか思いを馳せることになります.その後見立てて生成された画像がレーザーパターン上にビデオプロジェクタでプロジェクションマッピングされ重ね合わされて,観賞者はそれぞれの対応関係を確認します.一通り体験が終わると,別のポジションにレーザーが移動し,異なるパターンで上記体験が何度も繰り返されます.


September 20th, 2023.


Scott Allen