Nodal point of thought 2024-02-01
get-together
What are you doing these days? I answered "Digital Democracy" this time because "Plurality" doesn't convey the same message to the question "What are you doing these days?
(Aside from the fact that digital democracy is not something you "do".)
Then I'd be like, "What? What do you mean?" so it was a good opportunity to talk about a lot of things.
The Ministry of Digital Development (Ministry of Number Development) was created in Taiwan in 2022, and Audrey Tang was appointed as the first director.
There is a CC-0 book that Audrey Tang and Glen Weyl are writing on Github.
It is commonly known as the Plurality book
I think the Japanese should read more about this content because it's interesting.
Audrey Tan's lecture last December (I'm adding subtitles)
Glenn Weil, who works with Audrey Tan on Plurality (digital democracy), gave a talk when he visited Japan in January (I subtitled this one too!).
Chit Chat Memo
The spread of marijuana
If marijuana makes you subjectively happy, then optimizing for the purpose of "maximizing happiness" would mean that smoking marijuana is rational.
Even if there were no health hazard as an individual, it would be reasonable for a country to ban it because if its citizens are happy to smoke drugs and not work, their productivity will go down.
That's assuming that the productivity of the people's labor is important to the country, but in a world where AI has advanced to the point where it does labor instead, the assumption doesn't hold true.
As the demand for labor declines, the order of people with low labor productivity becomes "work is negative, so take a break".
As for "the elderly who don't work anymore," we still can't base that on labor productivity.
Urban high schools are rife with drugs and girls in your class buying pot.
Scary!
But there's an aspect of vaccine administration that comes from the circulation of stories like this.
People coming out of the countryside to the city for college are in trouble because they haven't been vaccinated.
It's also bad enough that contracts become available at age 18.
People who learned in the city return to the countryside to do network business.
Ducking by taking advantage of the perception that "someone back from the city has a money-making story to tell."
network business
It was the old system of buying and sending a fixed amount money order.
It's still there now, it's still there forever.
If the value of money is drastically reduced by AI and human relationships become more valuable, the model of complete transformation of human relationships will no longer be viable.
Let AI destroy capitalism and there will be no more fraud! ()
Happy!
This was not done in the existing paper and box voting system simply because it was too difficult to implement a different number of votes for each individual. Anyone can participate" and "give competent people more say" go hand in hand.
There is no compatibility between "equal say for everyone" and "giving more say to those who are capable."
The collective decision-making algorithm of "one person votes and chooses the candidate with the most votes" (majority rule) is a piece of crap and should be replaced with a better algorithm. Successful engineers, successful people in a meritocratic society, so they would be more receptive to a meritocratic, performance-based story.
That's right.
Updating "one-person, one-vote democracy" with the phrase "new democracy" means, of course, reducing someone's voice and giving it to someone else.
They're going to reduce the voice of the elderly, who have little remaining life expectancy, and give it to the child-rearing generation.
People basically rebel when they are about to be deprived, so it's better to focus on adding new ones.
For example, we could create a budget to support child-rearing, and then have a direct democracy where people with minor children are given a vote proportional to the number of children they have in order to make decisions on how to spend the budget.
The idea that "not voting is wrong" is an old-fashioned assumption.
I don't vote because the agenda on the ballot is not an agenda close to my heart.
If it were a more familiar agenda, people would be more willing to voice their opinions.
It is up to each individual to decide how big an agenda they want to participate in.
Not everyone needs to participate in national politics.
Many people are more interested in discussions on the scale of cities, districts, etc.
If people want to discuss the color of the roof of the bike shed, they should be free to discuss it.
Whether AI will govern humans
This question is a bad question that only vaguely captures "governance" in the first place.
AI executes as per the rules set forth.
The apparent that this is just DX, and the transfer of human behavior to AI will continue.
A human legislative body sets the rules.
There are probably a lot of people who would consider this place OK if they hadn't given it away.
I'm sure AI will eventually make suggestions to human politicians.
Self-driving cars and insurance
Even if self-driving cars are statistically less likely to have accidents than humans and are more reasonable, they won't be widely used because neither car companies nor AI companies can take the risk of accidental deaths occurring.
No, if there are statistically fewer accidents, then the insurance company just goes in and shoulders the risk.
Even if that solves the problem financially, it doesn't solve the emotional problem.
The issue of sentiments is ultimately subjective to the individual, so it's better to solve it with propaganda or something.
Like "technophobes who are emotionally repulsed by self-driving cars" and "harmful behavior that hinders the development of society."
Even if Japan is a little behind, as more and more self-driving cars come into use in China and the U.S., there will be a debate about whether it's okay to be left behind.
It was already out.
The story that OpenAI's Sam Altman is a hardcore basic income guy (commentary on his blog post from 2021).
Decreasing value of existing capital
Productivity gains from AI will cause the total wealth created by AI to overwhelm existing wealth.
r is return on capital, g is economic growth rate
But this is talking about an "economic growth rate that has been done by humans" of 1-2% per year.
In areas where AI is easy to apply, "productivity gains from applying AI" are much larger than r and g
The difference between r and g becomes irrelevant.
Why don't you buy stock in a power company?
Will power bottlenecks slow AI growth?
Isn't that why Sam Altman is investing in fusion power?
Will there be a monopoly by OpenAI?
Faction that does not happen
Data on the Internet is taken up and does not lead to differentiators.
We're going to get new data, and this is where the diverse players should be swarming.
Occurrence Faction (of the LDP)
In the process of going for new data, the most well-known player eventually gets the upper hand in negotiations.
Data does not have much weight in AI's ability to deliver value to customers, and OpenAI's internal know-how is more important.
The LLMs that are released to the world are a bit dumbed down by RLHF, if they are used by limited users for internal use, the limitations can be removed, OpenAI employees must be using the "smarter GPT4.5" without limitations on a daily basis.
Subsequent tweets
The story of a cash purchase of a townhouse with affiliate income in the hundreds of millions of dollars.
I did it because I had a service that originally said, "I've made so many sales as an affiliate," and I decided that if I wrote the program, I could do better, and even if I could take about 20% of that away, it was worth the time investment.
Unexpectedly, it got bigger.
The implementation is just static HTML generation from TSVs.
It's a composition where a non-technical stranger has validated the existence of a customer, and then scratched it with his technical skills... good.
Men my age, foundation and concealer on dark circles.
I didn't know that's how it works - I never thought about it.
Dentist recommends regular checkups.
---
This page is auto-translated from /nishio/思考の結節点2024-02-01 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.