PhD in mathematical logic, got into applied category theory lately.
Paul Z W
Karma: 1
Setting aside the concrete example of Pause AI (haven’t given it enough thought), I totally agree with the statement in the title.
Also, if I may: to some extent, you can accomplish things even when your soldiers aren’t as smart, or as ideologically aligned with you, as your scouts; same thing holds for officers. The historical example that comes to mind is the army of the Soviet Union: for some years at least, an important fraction of the officers were former officers of the imperial army; they were called “voenspetsy”, which means “military specialists”.
From the Wikipedia page on the Red Army:
“In June 1918, Leon Trotsky abolished workers’ control over the Red Army, replacing the election of officers with traditional army hierarchies and criminalizing dissent with the death penalty. Simultaneously, Trotsky carried out a mass recruitment of officers from the old Imperial Russian Army, who were employed as military advisors (voenspetsy).[19][20] The Bolsheviks occasionally enforced the loyalty of such recruits by holding their families as hostages.[21][page needed] As a result of this initiative, in 1918, 75% of the officers were former tsarists.[22] By mid-August 1920 the Red Army’s former tsarist personnel included 48,000 officers, 10,300 administrators, and 214,000 non-commissioned officers.[23] When the civil war ended in 1922, ex-tsarists constituted 83% of the Red Army’s divisional and corps commanders.”
Hi!
Have you heard of the ModelCollab and CatColab projects ? It seems that there is an interesting overlap with what you want to do!
More generally, people at the Topos Institute work on related questions, of collaborative modelling and collective intelligence:
https://topos.institute/work/collective-intelligence/
https://topos.institute/work/collaborative-modelling/
https://www.localcharts.org/t/positive-impact-of-algebraicjulia/6643
There’s a website for sharing world-modelling ideas, run by Owen Lynch (who works at Topos UK)
https://www.localcharts.org/t/localcharts-is-live/5714
For instance, they have a paper on task-delegation:Their work uses somewhat advanced maths, but I think it is justified by the ambition: to develop general tools for creating and combining models. They seem to make an effort to popularise these, so that non-mathematicians can get something out of their work.
Interesting ideas!
A hypothesis I found relevant to this phenomenon, similar to yours:
The problem “maximize impact per resources spent” is not well-defined a priori.
For instance, it depends on the time frame and scale: there could be very cost-effective smallish interventions, that can’t scale that much, versus very large scale interventions that require massive coordination, investment, “stubborness”, etc.
[Of course, you should try to see if such things actually exist in the real world; FWIW, I suspect they do]
It also depends on the entity you consider: is it you as an individual? The small group of people who are willing to listen and do a project with you? The whole EA community? Humanity?
You might be able to build a coherent system that takes into account these various levels though.
Another remark, that has more to do with execution than general principles, which you also touch upon: sharing all the information you have is not always a good idea. Unfortunately, the possible fixes (restricting information access to trusted people/groups) seem to go against the [EA/rationalist/...] culture of truth-seeking, open communication, etc.