Weekly journal 2023-09-02~2023-09-10

prev Weekly journal 2023-08-28~2023-09-02

• It’s been 5 days, but I feel like my brain hasn’t caught up with all the organizing, so I’ve switched.

- Many projects and activities were worked on during the week of August 28-September 2, 2023. - Numerous discussions were held on the development of AI-based pages and vector searches. - Aug. 29: Revisited the disregard for vector searches on AI-generated pages and considered whether the AI output should be included in the read. - August 31: Numerous activities took place, which were associated with several projects and pages. - September 1: Discussions centered on topics related to how to interact with and use AI, and also highlighted the challenge of balancing speculation and development. - September 2: The new concept of "multi-head thinking" was clearly verbalized, and this was recorded as a notable event.

Throughout the series of activities, it is clear that the focus was on discussion and experimentation regarding the role and appropriate use of AI and how to deal with its limitations. In particular, there was much discussion about the usefulness and limitations of features related to AI-generated pages and vector searches, and how they should be balanced.

• Intellectual Productivity Study Group by LLM, which has begun to compile a page on

Originally, the ability to develop a page by specifying a URL was implemented as an experiment in the development process, but we have come to feel that it is very useful in practical use.

8/29 Lift Ignore from Vector Search on AI-generated pages 8/30 Extending the Red Link with AI 8/31 Main branch stopped 8/31 Questions encourage verbalization, but there are different kinds of questions. 8/31 Introduction to ENCHI 8/31 Clarification of AI’s role is important. 8/31 Update Interval of AI Notes 8/31 A case study of SF prototyping for a junior high school student’s work experience 9/1 Page as a fluid process 9/1 Trade-off between speculation and development # distress 9/1 What is the role of AI in this project? 9/1 Is there an AI with multiple personalities? 9/1 Why not specify the purpose of each page of the AI note? 9/1 It’s buried at the bottom of the AI page. # Cause of distress 9/1 Discovering connections between different content 9/1 Summon other people’s AI to your diary

9/2 Multi-head thinking I feel this verbalization is epoch.

What was the specific activity behind the "Lifting the Ignore from Vector Search of AI-generated pages" on 8/29? What were the issues and objectives behind it? - Initially, it was designed, unknowingly and uncritically, to "not target the AI's output for reading." - And that's partly because we didn't have [[page memory]] in the beginning. - As the number of pages of AI output increased, the question arose, "Is it appropriate not to include this in the reading?" question arose. - They were identified by whether or not they had a 🤖 mark on the title. - For example, the human writing inspired by the AI's description is a useful fragment, but it's not the subject of the reading. - [[Extending the Red Link with AI]] More and more AI submissions without the 🤖 mark in the form of - Once removed. - Interference and convergence between [[multi-head]] branches now occurs - From this observation, "[[Multi-head thinking]]" was verbalized.

What was the background and purpose behind your efforts to “extend the red link by AI” on 8/30?

• see Extending the Red Link with AI

Numerous activities are listed on 8/31. Was there any particular major event or meeting on this date? - 🌀Where should I write? and collected those fragments as the thoughts were dispersed across a number of projects and pages. - The place influences the content. - If you think you’ve written something useful, you can immediately port it over to your personal project. - If fragments are collected, they can be tied together in a vector search.

What do you mean by “introduction to ENCHI”? Can you give us some specifics and objectives?

• A story about introducing AI to enchi, a reader community of engineers’ intellectual production techniques.

9/1 on “Tradeoffs between Contemplation and Development” tagged as “Painful”. What have been the challenges or difficulties in this topic?

• There was a lot of speculation and no development.
• I struggled with whether I should continue contemplating or stop and proceed with development.

Topics such as “Update Interval of AI Notes” and “Buried at the Bottom of AI Pages” appear to be discussions and considerations about how to interact with AI and how to use AI, can you give us some specifics, background, and objectives?

• AI notes were initially updated once a day. With the introduction of multi-heads, 22 were made per day at one time, which is a burden for humans, so a cap was set.
• However, this cap makes it impossible to predict when a page will be subject to AI updates, and we thought that might not be a good idea.
• The buried issue also relates to putting a 🤖 mark in the title.
  • The problem of human writing sinking to the bottom of 🤖 pages and not being found.

Regarding the “multi-head thinking” on 9/2, you say “I feel this verbalization is an EPOCH”, what exactly do you mean by this “EPOCH”? Also, what were the key points or important findings of the day’s activities and discussions?

• I think it’s an important event that the unknown concept that we’ve been “trying” to do for a while now has a name: “multi-head thinking”!

---

This page is auto-translated from /nishio/週記2023-09-02~2023-09-10 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.