# Creating Monochrome Digital Paintings out of ChatGPT

Canonical URL: https://jeanyoon.ch/oi/texts/creating-monochrome-digital-paintings-out-of-chatgpt
Type: essay / research note
Author: Jeanyoon Choi (최정윤)
Date: 2024-03-23

## Summary

This 2024 note proposes an interactive monochrome made from accumulated text generated by ChatGPT or another language model. Drawing on Dansaekhwa and on the principle that the sum of parts is not the whole, also central to his work ≠, Jeanyoon Choi imagines a keyword typed on a keyboard producing text across six screens that accelerates and overlaps until words become illegible. The aim is to treat language as material, letting participants move from searching for meaning to affirming the signifier.

## 요약 (Korean)

최정윤은 단색화에서 영감을 받아 ChatGPT가 생성한 텍스트를 여러 화면에 겹겹이 쌓아 의미가 사라지고 물질이 되는 인터랙티브 모노크롬을 구상한다.

## Concepts defined

- **ΣPart ≠ Whole**: Choi's principle that accumulating parts can change their character, so logical, interpretable individual outputs may combine into something chaotic and illegible: a qualitative rather than quantitative change.

## Questions this answers

### Can AI-generated text become an abstract painting?

In 'Creating Monochrome Digital Paintings out of ChatGPT', Jeanyoon Choi proposes layering language-model text across screens until individual words become illegible, so ChatGPT becomes a source of material rather than a voice to be understood, producing a monochrome-like surface.

### How does Dansaekhwa relate to Jeanyoon Choi's work with language models?

Choi draws on Dansaekhwa's repetition, materiality and abstraction, contrasting it with descriptive language, and asks whether accumulating descriptive AI text on a massive scale can turn meaning into material.

Topics: dansaekhwa, monochrome, large language models, text as material, abstraction, interactive installation

References: [ChatGPT](https://en.wikipedia.org/wiki/ChatGPT), [Dansaekhwa](https://en.wikipedia.org/wiki/Dansaekhwa), ≠ (artwork by Jeanyoon Choi), Laboratory Occupied (Jeanyoon Choi), [Glenn Ligon](https://en.wikipedia.org/wiki/Glenn_Ligon), [Jenny Holzer](https://en.wikipedia.org/wiki/Jenny_Holzer)

## Full text

An interactive monochrome made from language: layers upon layers of text generated by ChatGPT or another language model.

I am thinking of Dansaekhwa, of repetition and materiality, and of abstraction pushed towards its limit. Language appears to occupy the opposite position. We ordinarily use it to describe. Poetry can make language abstract, but that is far from its everyday use—and from much of the descriptive, repetitive writing produced by ChatGPT. My question is what happens when that language is treated as material.

What happens when descriptive, computational or supposedly intelligible outputs are repeated and accumulated on a massive scale? I keep returning to ΣPart ≠ Whole. Accumulation can change the character of its parts. It is a qualitative transformation, not simply an increase in quantity.

This is also the concern of 『≠』. Individual algorithmic outputs may appear logical and interpretable, yet their combination produces something chaotic and difficult to read. The apparent equality of each part does not guarantee an intelligible whole.

Could the same transformation produce a monochrome? I imagine returning to some of the visual experiments of Laboratory Occupied. Text begins with meaning, but as it accumulates, it becomes illegible. It is consumed as material. ChatGPT itself becomes a source of material rather than a voice we must understand. Does an infinite accumulation of meaning become an absence of meaning?

One possible arrangement uses a keyboard as the input and six screens as outputs. A participant enters a keyword; language models generate text across the screens. The context of those texts remains to be decided. As interaction intensifies, their accumulation accelerates until overlapping layers fill the screens and individual words can no longer be read.

Glenn Ligon and Jenny Holzer are relevant references, although the disappearance of meaning is not necessarily their purpose.

At first, the participant may cling to what the words mean. Later, could the interaction release them from that demand? Could they affirm the signifier—the letters, the marks, the equivalent of ink—rather than continually searching for the signified? They would enter a larger system produced by language without needing to decode every sentence.

This is one possible meaning of immersion for me. The movement from a blank screen to an accelerating monochrome matters as much as the final surface. Interaction carries the participant from meaning to its dissolution, from language to colour, from description to abstraction. The resulting surface continues to change, even when it is no longer clear exactly where a change has occurred.

---

Jeanyoon Choi (최정윤) is a Korean computational artist and researcher (b. 1999), PhD candidate at KAIST XD Lab, making multi-device web artworks. Profile: https://jeanyoon.ch/oi/en. Index: https://jeanyoon.ch/llms.txt.