from Experiments to make GPT4 compensate for context Good abstraction and bad abstraction https://twitter.com/nishio/status/1677861710789775367?s=20

nishio I interpreted “A is A” as the subtitle of the third part of the Atlas of Shrugging Shoulders, and “The Art of the Betting Booth” as drawing out private investment, am I correct?

_tea_two Interesting w yes, mine is correct. I was using it to separate it from the general usage and to limit it.

_tea_two Since we’re at it (?) I’d like to add a few more “intuitions” to help cultivate ChaGPT.

I had an underlying theme of what is a good abstraction and what is a bad abstraction.

I interpreted moto sg’s two tweets as taking issue with the process in the “concrete → abstract A → abstract B” structure, especially the “abstract A → abstract B” process.

_tea_two So what is a good abstract B? About. This time, I interpreted that the criterion for a good abstract is “whether it is based on concrete”, and brought up “A is A” as a verification tool.

  1. abstract B has not deviated from the original concrete of abstract A
  2. abstract B is an association from abstract A. The original concrete does not apply, but it can be applied to another concrete.

_tea_two I think this story is about 1 in particular being preferable, but “A is A” itself can be used for 2 as well. That is all.


This page is auto-translated from /nishio/良い抽象と悪い抽象 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.