1) Thanks for the pointer to the data, I have to agree that if the surveys are representative of EA / rationalist community, than actually there are enough medium sized hubs. When plotting it, the data seem to look reasonably power-lawy - (an argument for a greater centralization could have the form of arguing for a different exponent).
I’m unsure about what the data actually show—at least my intuitive impression is much more activity is going on in Bay area than suggested by the surveys. A possible reason may be the surveys count equally everybody above some relatively low level of engagement (willingness to fill a survey), and if we had data weighted by engagement/work effort/… it would look very different.
If the complains that hubs are “sucking in” the most active people from smaller hubs, than big differences between “population size” and “results produced” can be a consequence (effectively wasting the potential of some medium sized hubs, because some key core people left, damaging the local social structure of the hub)
2) Yes there are many effects leading to power laws (and influencing their exponents). In my opinion, rather than trying to argue from the first principles which of these effects are good and bad, it may be more useful to find comparable examples (e.g. of young research fields, or successful social movements), and compare their structures. My feel is rationality/EA/AI safety communities are getting it somewhat wrong.
Certain ‘jobs’ seem to have this property: a technical AI researcher in (say) Japan probably can have greater EV working in an existing group (most of which are in the bay) rather than trying to seed a new AI safety group in Japan.
This certainly seems to be the prevalent intuition in the field, based on EV guesstimates, etc., and IMO could be wrong. Or, speculation, possibly isn’t wrong _per se_, but does not take into account that people want to be in the most prestigious places and groups anyway, and already include this on an S1 level. And this model / meme pushes them away from good decisions.
1) Thanks for the pointer to the data, I have to agree that if the surveys are representative of EA / rationalist community, than actually there are enough medium sized hubs. When plotting it, the data seem to look reasonably power-lawy - (an argument for a greater centralization could have the form of arguing for a different exponent).
I’m unsure about what the data actually show—at least my intuitive impression is much more activity is going on in Bay area than suggested by the surveys. A possible reason may be the surveys count equally everybody above some relatively low level of engagement (willingness to fill a survey), and if we had data weighted by engagement/work effort/… it would look very different.
If the complains that hubs are “sucking in” the most active people from smaller hubs, than big differences between “population size” and “results produced” can be a consequence (effectively wasting the potential of some medium sized hubs, because some key core people left, damaging the local social structure of the hub)
2) Yes there are many effects leading to power laws (and influencing their exponents). In my opinion, rather than trying to argue from the first principles which of these effects are good and bad, it may be more useful to find comparable examples (e.g. of young research fields, or successful social movements), and compare their structures. My feel is rationality/EA/AI safety communities are getting it somewhat wrong.
Certain ‘jobs’ seem to have this property: a technical AI researcher in (say) Japan probably can have greater EV working in an existing group (most of which are in the bay) rather than trying to seed a new AI safety group in Japan.
This certainly seems to be the prevalent intuition in the field, based on EV guesstimates, etc., and IMO could be wrong. Or, speculation, possibly isn’t wrong _per se_, but does not take into account that people want to be in the most prestigious places and groups anyway, and already include this on an S1 level. And this model / meme pushes them away from good decisions.