Yesterday and today I was working on the Plurality index.

  • Jobs that let AI do the work

Do you have anything to share?” I had a conversation about how tasks that can be easily directed and handed off can be easily prompted to LLMs to get results “without the unpredictable wait times that occur with humans”, and can be “psychologically safe” in the sense that you don’t have to worry about social risk if the results don’t meet expectations, and can be ruined or discarded as many times as you want. You can make them start over and discard the results without explanation.

I was reminded that as LLM develops, the shape of work will probably change.

nishio @hal_sk @0xtkgshn If you want your name to appear as a contributor in the English print edition of PluralityBook, Monday is the deadline (for installing Gov4Git, applying to join, and receiving credit for your contributions).

0xtkgshn too much hassle

nishio Well, there’s no benefit other than getting it in the paper, so you can go through with it if it’s too much trouble.

nishio I thought until this morning that it’s just a big deal because it’s on the paper, but I’m so sorry to see my name next to Vitalik’s. I’m so glad to see my name next to Vitalik’s. image

pPluralityBook claude.iconThe main flow on April 10, 2024 is as follows

  • Jason Entenmann shared the latest docx file with nishio via SharePoint. nishio successfully converted it to PDF and downloaded it.

  • nishio asks Jason Entenmann to change his name from “Nishio Hirokazu” to “NISHIO Hirokazu”, which is approved.

  • nishio uses GPT-4 to generate chapter and page matches from the most recent table of contents data.

  • nishio fixed a problem with the PDF to JSON conversion script, reducing the number of “missing keywords” from 242 to 79.

  • An index of 1,852 entries was created using Claude 3 Opus, with Claude providing comments on the contents.

  • nishio created a pull request on GitHub, announcing that anyone can now review and pull index_for_manual_edit.tsv after the merge.

  • NISHIO completed the day’s work and informed the group that it would resume the next day (16 hours later).

  • Diary 2024-04-09 ← Diary 2024-04-10 → Diary 2024-04-11 100 days ago Diary 2024-01-01. 1 year ago Diary 2023-04-10.


This page is auto-translated from /nishio/旄蚘2024-04-10 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.