I previously expressed opinions that in order for an AI pause movement to succeed, multiple large political groups in the US including christians, communists, liberals, libertarians etc will need to form a broad coalition against AI
Each group can upvote or literally vote for their own leaders. But then these leaders will have to coordinate each other. Or one leader can earn votes from multiple groups at the same time, although this actually really hard for a leader to do.
I am now starting to think that maybe I should not be willing to ally with communists in particular.
It seems quite likely that if we manage to elect a US president that is truly anti-AI to the extent required (actually shut down the AI companies, confiscate large GPU clusters in your own country, actually sit and negotiate with China, etc), then that person won't just be anti-AI but also anti- a number of other things. They might be anti-silicon valley more broadly. They might be anti-technological progress more broadly. They might be anti-businesses more broadly.
My view used to be that I was basically fine with this, and that this is all acceptable damage as long as they do deliver on being as anti-AI as actually required. I was mostly a lot more concerned that they won't be as actually anti-AI as required, and the people who could have played that role instead would fail to acquire enough power to get there.
If I take this thinking to its logical extreme however, I am basically saying I am fine with literally getting Stalin in power so that he can turn the US into a failed communist dictatorship, with purges and mass starvation and so on.
And, this actually gives me pause? I am not sure if I endorse this or not.
From a naive utilitarian lens, I can maybe justify it.
A
I currently assign ~8% probability of extinction ~8% probability of permanent global dictatorship if I do nothing.
Assuming 8 billion people, that is already 640 million deaths, plus another 640 million people losing atleast a few QALYs each.
And then you have to add up all the gazillion future minds who either weren't created at all, or were created in a dictatorship possibly involving suffering
B
Imagine that I was in a position where me endorsing a communist politician increased their likelihood of being elected by 33% (absolute, not relative).
Assume the probability this politician will go full Stalin is also 33%.
Assuming they run the entire US sphere (maybe 3 billion people)
Assume they will kill 100 million people, and reduce atleast a few QALYs for these 3 billion people
This now adds up to 10 million deaths, and few QALYs reduced for 300 million people.
Here you will have to use a significant probability, that Stalin's successors might not be the same as him, and the future will be better. You have to also assign significant probability that maybe the people after him too will fail to build democratic govts, and one of these govts will eventually build ASI too
I am not a naive utilitarian. The fact that the total numbers for A and B are within same order of magnitude itself tells me that this is maybe a bad gamble to be taking.
If I were Yudkowsky-level doomer, then sure, getting Stalin in power to delay ASI by even another 5 years or 10 years is obviously a good thing. If you believed in 99% doom in next 10 years or similar, then literally any plan looks like a better plan than doing nothing about it. I am not that doomer yet, however.
Subscribe
Enter email or phone number to subscribe. You will receive atmost one update per month