Shutaro Aoyama

Search

SearchSearch

What is tkgshn

Nov 20, 2024, 8 min read

  • It seems like the following content is a bit mixed up with two different topics:
  • tkgshn, it is a fact that foundational subjects commonly taught in schools, such as mathematics and humanities, are very unstable.
  • Despite this instability, they have been producing results in related fields and continue to do so. Why is that?
  • Hypothesis: It is due to the Positive Feedback Loop below:
    • ① Gathering information from external sources -> ② Identifying a position -> ③ Taking a position -> Gathering more information than ①
      • Agreement (tkgshn)
    • The impression is that this cycle is continuously active due to high abilities in ①②③.
    • ①: Gathering information
      • Ability to connect with people, listen to discussions, borrow others’ knowledge, or conduct surveys independently.
      • I particularly felt this ability while watching a Zoom discussion about Decartography with many acquaintances.
    • ②: Identifying a position
    • ③: Taking a position
      • i.e., Twitter influencer power
      • By excelling in this, higher-quality information tends to gather in ①.
  • Instead of the approach of logically stacking blocks one by one on the foundation, it gives an impression of a probabilistic approach that quickly refines accuracy by running a Feedback Loop based on information, similar to machine learning.
    • tkgshn = Large Language Model theory(?)
      • March 17, 2023: As the use of LLM has become more feasible, I have started to understand this sensation quite well (tkgshn)- It’s like throwing all objects into a vector space and then thinking about it, kind of like that, w(tkgshn).
      • The context is a bit different though, w(tkgshn).
  • This was also brought up a bit during the end of the year SGG, right? w(kota-yata)
    • What kind of feeling is that? w(tkgshn)(tkgshn)(tkgshn)
  • Doing things like Hallucination, strength in language generation, various phenomena seem similar (blu3mo)

Is it possible to extract only the good parts and turn them into insights that everyone can use? (tkgshn)

  • Is it similar to /tkgshn-private/Life_Strategy?

  • Rather, it seems like it could be useful to utilize LLM to interact with society, right? (blu3mo)

    • Letting LLM say random things for now? w(tkgshn)
    • Giving LLM a position, something like that (blu3mo)
      • Is that a way to make LLM smarter? Or a way for humans to become smarter? (tkgshn)
        • Closer to the former (blu3mo)
        • Like a way to make LLM thrive in human society (blu3mo)
        • Continuously having ChatGPT say incorrect things and creating a situation where human experts want to provide feedback on it? (tkgshn)
          • But isn’t that what everyone is criticizing on Twitter right now? (tkgshn)
          • Is it crucial to point out flaws? (tkgshn)
            • Cunningham’s Law

            • “The best way to get the right answer on the internet is not to ask a question, it’s to post the wrong answer.”

          • Deliberately doing this is smart but also seems malicious (blu3mo)
        • Well, that’s one way, but it seems like there could be more at a higher level (blu3mo)
        • ① Gathering information from external sources -> ② Finding a position -> ③ Taking a position -> Looping back to gather more information than ①, that kind of concept (blu3mo)
          • It seems possible to have them connect with people, listen to conversations, borrow people’s thoughts and knowledge, etc.
            • tkgshn seems to create situations where both parties benefit from engaging in interesting conversations
          • The structure is the same, so I’d like it to be verbalized as “at a higher level” (tkgshn)
            • Cunningham’s Law, for example, takes advantage of spreading incorrect information to create disadvantages or negative emotions in others
              • It seems like there’s no positive externality (blu3mo)
            • Should we just avoid spreading incorrect information? (tkgshn)
              • That’s too simplistic (blu3mo)

  • “Why is blu3mo with (tkgshn) right now?”

    • It’s fun
    • Could an LLM of this Takagi learning model be created by making a bot that is fun to talk to?

Does blu3mo observe my actions and take them into account? If so, what is it? (tkgshn)

  • The evaluation mechanism is buggy

Concrete example:

  • “I put bluemo’s bag on (tkgshn)‘s bed and tried to bring it into the hostel to choose luggage.”

  • Actions:

    • First, I handed bluemo the card in front of the staff to see if they would let us pass
    • → We were stopped
    • → Negotiated
      • Just for a moment, isn’t it okay?
    • → Negotiation failed
    • → Waited normally in front
      • Since we didn’t pay any costs, we just explored the “logically possible range” (tkgshn)

  • Thoughts:

    • Try it out, there’s no downside. If it works, there’s only an upside. Why not do it? (tkgshn)- Upside

  • Downside

  • It’s very rational and makes sense, but I think the majority worries about being scolded (blu3mo)

  • Children, due to differences in physical size and economic dependence on adults, it makes sense for them to be concerned about being scolded. However, there comes a time when they need to become independent, and during that process, their cognitive structure needs to change. But many people do not change in that process (nishio)

  • It’s better to ask the same thing to people multiple times

  • [Attack Power], Contacting Unknown Influential Figures

    • I feel like I’ve been able to do this somewhat recently (blu3mo)
    • However, I think most people face emotional barriers to this rational behavior
    • Why do you think there are emotional barriers? (tkgshn)

  • Skills to come up with rational choices

    • Do all the options in the Logical Space of Choices (tkgshn)
      • It’s not exhaustive, but it seems like it requires a high level of exploration ability (blu3mo)

  • Skills to ignore irrational emotional blockers

  • [Decision-Making Ignoring Frameworks]

Concrete Example 2

  • After enjoying Devcon in Bogota, I ended up in New York even though I didn’t have a home. (I had a B-1 visa, so I didn’t need to book a return flight) (tkgshn)

    • The reason was, “America seems fun?” (tkgshn)
      • “Because everyone was saying, ‘It’s better to go to America.‘”
    • By the way, European Triumph Spring 2023 was for the same reason. lol (tkgshn)
      • A friend suggested, “Since Public Goods aren’t really accepted, how about Berlin?”
    • Reasons for consideration
      • Simply put, no liabilities. Light on my feet
      • Reducing sunk costs (tkgshn)

  • It might not be about tolerating uncertainty, but rather feeling strong about being certain

    • [Ability to Go for It]

When I talked with tkgshn in April compared to when I talked with three people the day before yesterday, my observations improved, so I feel like language articulation might progress upon further consideration (nishio)

  • My observations improved

  • Did your observations about “tkgshn from nishio’s perspective” improve? (tkgshn)
    • Yes (nishio)
      • Difference Between Conversations with Two People and Three People
      • I’m beginning to see commonalities with a friend C
      • Friend C: Noticing a common aspect where he preferred reading the works written by Drucker himself during the Drucker boom rather than the popular novel versions, he encouraged me to enter the Tokyo Tech MOT as a working adult graduate student
      • tkgshn: Encouraged me to come to Plurality Tokyo after seeing Scrapbox
      • Afterwards: As our common knowledge base grew, it became easier for me to interpret and facilitate language articulation in different ways, resulting in positive externalities
    • Cases Where Knowledge Accumulation is Not Necessary
      • Is this also about “tkgshn from nishio’s perspective”? (blu3mo)
      • When I finished writing and came to paste it, it was already pasted lol (nishio)

Graph View

Backlinks