- Question by Daiki Teramoto on [Unexplored Junior OpenAI / ChatGPT Roundtable
I think this is Important Question.
(2023-04-04)Not sure if this is a good answer, but I’ll jot down what comes to mind.
- [The Art of Worldly Wisdom.
- There are “understandable”, “unintelligible”, “interesting”, and “uninteresting” for fragmentary sentences.
- Sometimes associating and connecting one thing to another in response to a sentence.
- For example, The wise man is immortal. was connected with Lao Tzu
- It’s interesting what these things occur.
- Past Hatena Diaries that have been imported and are searchable but not linked, Twitter and Facebook that have not even been imported, and PDFs of lectures that have been partially imported but the method is questionable.
- I would like to see these become “A better form of knowledge representation”.
- That “better form of knowledge representation” is not yet known.
- Vector search.
- Nodal Point of Thought 2023-04-04
- If you read them and say, “Answer the questions,” you will be told, “There is no information. If you read them and say, “Summarize,” you will be told, “There is a lot of information.
- I want you to “think.” What is “thinking?
- When I reinforce my thoughts in Kozaneba, I first chop them into smaller units.
- Scrapbox is long-term memory, vector search results are “associations.”
- Chop it up into small pieces and place them in space.
- What is spatial arrangement?
- Work close together with close associative distance - associative connection
- Whether it is due to human limitations to make it two-dimensional or whether it is inherently important is a mystery.
I wonder if I could read the book and help you make a good leverage memo.
- And I thought 2009 Leverage Memo Organizing.
- It doesn’t seem to make sense to just make a leverage memo without reading it.
- However, it might be a good idea to make a “level 1 leveraged memo” that is read by an AI first, and then cut out the muddled writing style and make bullet points, etc., and then a human can read it.
- You can skip to the original text if necessary.
- Image of AI writing reading notes in Scrapbox first.
- People write their thoughts on those bullet points.
- Do we need to leverage it further into a leverage memo?
- Maybe it’s important for humans to read back over time.
- Aside from whether or not the AI summarizes when you read it back.
- Maybe it’s important for humans to read back over time.
- However, it might be a good idea to make a “level 1 leveraged memo” that is read by an AI first, and then cut out the muddled writing style and make bullet points, etc., and then a human can read it.
2023/11/20
- I agree with the main idea.
- A: Conversion to Bullet Points
- Books and other texts for the general public are “formatted that way”.
- Not in the bulleted format that Scrapbox and Outliner users consider a “good format for intellectual production”.
- First, this conversion can be made generic.
- B: Subjective pickup
- Next, regarding picking out the important parts from the bullet points
- This varies from person to person because “what one perceives as important” is subjective.
- The model should be created with training data from one person with a strong personality (e.g. me).
- Mixing evaluators is no fun.
- C: Expand your thinking from the bullet points you picked up
- I think this place should be human.
- Maybe we could do a vector search and show it as a reference.
- The fact that the raw “vector search results” are not good enough.
- Search targets should also be A in advance.
- The fact that the raw “vector search results” are not good enough.
- Provide a link to the pickup.
- I can do this.
- If the link connects (as long as the link is not too large) it is important
This page is auto-translated from /nishio/自分が1000人居たら何をするか? 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.