• I thought Iā€™d ask ChatGPT about a specific question about the concept of ā€œitā€™s faster to do X.ā€
  • For example.
    • How much is 3/2 in Python? Itā€™s faster to try than to look for a description in a paper book.
    • When I get an error message, instead of looking for something in a book about the error message, it is faster to just take the ā€œunchanged partā€ of the error message and Google it. StackOverflow is usually a hit.
      • Rather than trying to translate the error message itself, it is better to search for the error message in English, find a good explanation on StackOverflow, and machine translate it.
    • On the other hand, certain questions are ā€œIt is faster to read the reference manual,
    • Certain questions become ā€œthis is faster to read the source code.
  • A seasoned engineer unconsciously picks and chooses the ones that look good from among many information acquisition options and uses them.
    • I realized that there is this ability to predict ā€œwhich option will return the answer you seek with what degree of probabilityā€.
      • ā€œThe ability to predict ā€œwhich option will return the answer you seek with what probabilityā€ is acquired through experience, so you donā€™t have much confidence in options with which you have little experience yet, Trade-offs between use and exploration.

context

  • I asked GPT4 about the rationale for the superiority of the barbell portfolio over the ladder portfolio, but he didnā€™t seem to understand much, so I thought, ā€œThis would be faster to read a book,ā€ he said.
  • Iā€™ve done some light research and Iā€™m wondering if it means that ā€œunder conditions where the interest rate on the loan is highly variable, a barbell that is stretched only at both ends performs better than a ladder that is stretched evenly across the entire loanā€ or ā€œthatā€™s because the ladder is not sufficiently diversified against the risk of interest rate fluctuationsā€. Iā€™m not sure if this interpretation is correct or not, but itā€™s quicker to read what he wrote.
  • Itā€™s not so much ā€œitā€™s faster to do itā€, but more like ā€œI donā€™t think I can google any moreā€.

https://twitter.com/nishio/status/1636919490247270401 https://twitter.com/nishio/status/1636971380645441536


This page is auto-translated from /nishio/Xć—ćŸę–¹ćŒę—©ć„ 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. Iā€™m very happy to spread my thought to non-Japanese readers.