epistemic status: quick stream-of-consciousness on things i’ve been thinking on lately
The notion of leverage is very good because it is the opposite of things being Lipschitz. When things are Lipschitz, bounded input perturbations beget bounded output perturbations.
Something is high-leverage when a small input perturbation can lead to very large downstream effects. Previously, I thought in terms of an injection into the global production function. Or a tiny amount of force applied in just the right place.
The problem with this conception is that it was very individualistic.
Instead, it makes sense to think about things more like this:
In fact, you need only induce the first plane because it will take care of inducing the latter/downstream plane. It makes no sense to think in terms of “what is the most I can accomplish with a tiny amount of force / 1 FTE?” because you need not limit yourself to this, especially in the age of coding agents.
It makes a lot more sense to think about setting off self-propelling processes.
Currently, my friend and I are organizing a high-leverage research hackathon. It should take <10 hours of organizer time and hopefully have a large impact.
A lot of mental reframing is important, e.g.:
“people are scared of traveling back in time and changing the present, yet no one believes the things they do today will matter”
At post-AGI political theory reading group1 last week, we chatted about how the most important thing is avoiding bad value lock-in. Oh — it’s psychologically unhealthy to attend to that which we reject, but ‘locking in good values’ isn’t the answer either; long reflection is good.
And I think we will need mass surveillance to combat VWH, but mass domestic surveillance is currently unpopular. We haven’t articulated forms of surveillance that respect all our priorities yet.
An ongoing project for me is to understand what (if any) design principles enable surveillance that enables safety, while balancing all parties' need for flourishing; and what leads it to fail, leading to authoritarianism?
Posts in pipeline
Replace ‘forecasting’ with ‘threading the needle'
Thinking in terms of life cycles helps. We are currently in a life cycle where our challenge is actually procuring enough vitae per person. This life cycle may not be indefinite (cf. population ethics, and btw all you need to know about population ethics is you have to take an anti-Malthusian-trap stance); you can choose which to focus on; you can make your work conditional on others’; you can decide what to make your problem. P.S. as soon as someone else articulates it it is solved. Probably don’t work on someone else’s agenda. We need more agendas. Pitch someone on starting a new program / team by articulating outcomes.