Relationship between LLM and Plurality
2023-05-19
Relationship between LLM and Plurality
LT at [LLM Meetup Tokyo 2
Since I have been focusing on Plurality-related issues for the past month (Plurality Tokyo - Manazuru 2023-05-13), I would like to talk about the relationship between LLM and Plurality. The concept proposed by Audrey Tang, Minister of Digital Agency of Taiwan, et al. Concerns that linking AI to government will increase the power of government
We need to consider using AI to make the people more powerful to balance the situation.
This direction is called "Plurality" using "Plural (plural)" as a synonym for "Singular (singular)" in the "Singularity" that is supposed to be brought about by AI.
Technical Components of Plurality
There are several, but it's impossible for LT to explain them all, so I'll just mention one.
Enables "broad listening," where millions of people can hear the essence extracted from the distribution of opinions of their peers, and can enhance democratic careful deliberation on a large scale. For example, if you have a decent discussion with 100 people on Twitter, if you wrote 1 tweet your opinion, of course you should read 99 tweets
Are you doing this? You're not doing it.
Why aren't we doing it? Because it's too much of a burden for flesh and blood.
Then let's support that with digital technology!
concrete implementation
Often used as a component Polis. Lower the burden of "listening to many people's opinions" by clustering "people with similar opinions."
This Polis does not use LLM.
Natural language processing is the future
This is an area where there could be interesting developments using LLM in the future!
Circumstances unique to Japan
Erin Meyer's research shows that Japanese culture is such that "it is better for everyone to build consensus than for the boss of an organization to make top-down decisions." Many want to "involve themselves in the decision-making process."
But many people don't speak for themselves.
You don't speak your mind, but you want your opinion to be taken into account.
Why Japanese People!
Need to design a system that takes care of this Japanese cultural background.
https://gyazo.com/3b5e969822d2773f6a36009707b6c2f6
Each person's statement is first heard by the LLM
Anonymize and summarize it and feed it back to everyone.
(If I had more time, I would have liked to get to the point of actually building and demonstrating this, but I couldn't make it in time.)
---
This page is auto-translated from /nishio/LLMとPluralityの関係 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.