This is a special post for quick takes by Morpheus. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
I wish there were more concrete examples of how people think about being virtuous TDT agents in practice on earth where people are only medium good at doing TDT. I want to know where they think they succeeded with that as well as times when this lead them astray because they didn't understand some cultural nuances, themselves, or something else. If others are interested in that I will think about examples in my personal life I can share.
I am not asking for abstract rules. I am asking for how people think they implement those abstract rules in practice.
3JBlack
TDT and related policy optimization rules are effectively impossible to properly apply in real life due to superexponential explosion on the set of inputs to the policy. Somewhat related rule of thumb approximations are the best you can actually get.
Brains exist. They benefit from well curated training data just as artificial minds do. Really, lesswrong has material on coordination spread all over the place, but almost nobody can read it all. If I have a hard time making sense of it, then I think there is more distillation to be done. For example, I was thinking about how you actually bootstrap yourself to dath ilan levels of coordination technology and in the process I accidentally stumbled over stuff that then I noticed is already in project lawful.