Specific UI concept:

  • Multiple columns where the text in each column can be collaboratively edited.

  • Editing in each column is transformed and synchronized by LLM.

    • Editing in Column A will be converted and applied in Column B as well.
  • What kind of transformations could be possible?

    • Various levels of granularity.
      • Imagining something like Fractal Summarization.
        • For instance, the main text and a three-line summary displayed side by side.
        • Editing the main text will reflect changes in the three-line summary, and vice versa.
    • Various languages.
      • Japanese text and English text displayed side by side.
        • Editing one will…
      • It seems very useful to have Japanese and English speakers collaborate.
        • It seems plausible but hasn’t been seen before.
    • Various topics.
      • Similar to asym-chat (Anjasshu State Chat).
      • Abstracting a concept and converting it to a different topic.
        • For example, looking at a problem in system design from the perspective of organizational design might spark solutions.
    • Various cultures.
      • Simulating how one’s current writing might be perceived by “Americans” or “the elderly,” for example.
        • By observing and adjusting from different perspectives, the original text can be refined.

Benefits:

  • Facilitating collaboration and mutual understanding across cultures.
    • —> This part is with Ari-sensei.
  • Even when used alone, it aids in conceptual manipulation, thinking, and writing.
    • It becomes an experience of examining and refining a text from various angles and granularities.
    • Using this for organizing thoughts or writing seems intriguing.
    • —> This part is for lab use.

Implementation:

  • Implementing synchronization processes involving LLM transformations for real-time collaborative editing may require some ingenuity.

  • How to apply transformation X in Column A to which line in Column B?

    • Maintain correspondence? Let LLM handle all applications smoothly?
  • How to resolve conflicts?

  • There’s a good chance that letting LLM handle all the complex tasks might lead to a satisfactory solution (blu3mo).

  • If spatial and experiential information can be converted into language, this mechanism could be applied.

    • Imagining the processing of space being reduced to language processing.