Ramez Naam and Alex Imas argue AI is dispersing power like electricity rather than centralizing it like social media
Open-source models help citizens navigate laws and monitor officials.
Very good from Andy. I think I agree broadly, especially with point 4. Accord confusion right now is to what extent AI is going to be able to eat the entire world versus make the entire world more productive. The labs are clearly trying to eat the world and if this exceed then yes they can internalize a large chunk of the current usable world output.
However if hayekian knowledge is true and there exists enough local complexities or implementation details or or requisite ability to withstand unseen unknown unknowns all of a sudden this becomes one where you can still have a few very large companies but it is not going to be a threat per se.
The other core component is that this presumes the labs can will continue put models out for everybody to use in some way or another. Or that open source models will give us a lot more freedom. I don't know if either of those things are necessarily true but it is a double edged sword if they don't because if they don't they don't get the data to make the next model better either.
Here are 4 reasons I'm not convinced AI is going to concentrate power and fuel authoritarianism, even though it might. (1) AI is competitive and diffuse. Open-source models are proving highly capable. This means no one model controls what we can and can't do, because we have options. (2) AI may empower citizens, not just states. States can use AI to process surveillance data, yes. But citizens are also using it to get smarter about the legal process, monitor elected officials, communicate more effectively, etc. (3) AI may not solve the principal-agent problem that dictators face with their enforcers. Is it so easy for an authoritarian to be sure his AI answers only to him? I'm not sure. (4) (meta point) We are terrible at predicting the future of stuff like this. Remember when everyone thought social media was great for democracy? We should spend a lot more time thinking about and studying this---and that's what I'm planning to do! I've attended a few very interesting conferences recently on AI and the concentration of power. They've convinced me that it's an important topic. But I'm not so convinced that the conventional wisdom in these circles---that AI is essentially and inevitably a pro-authoritarian technology---is right.