Richard Ngo
@RichardMCNgo
studying AI and trust. ex @openai/@googledeepmind, now thinking in public.
I increasingly believe that there are fundamental principles which simultaneously govern the designs of well-functioning minds, organizations and societies. Once we pin them down with mathematical precision, we’ll understand the world more deeply than we can currently imagine.
Morality as a power-sharing compromise, as below, is more true than utilitarianism. But it’s not fully true, because we’ll discover game-theoretic principles (e.g. bargains across counterfactual worlds) which motivate “cooperation” even with beings that lack power here and now.
Utilitarians, please listen: Utilitarianism isn’t real, it’s just a useful compromise to make sometimes when humans have to get together to decide on rules: “I care about my family the most, and you care about your family the most, so while we’re working together let’s just…
Are there any tools that calculate PageRank-style scores for scientific papers, social media accounts, etc? I.e. paper citation counts weighted by how many citations the citer has, twitter follower counts weighted by how many followers of their own each follower has, and so on.
“Ask 2 clarifying questions before contradicting someone” is an AMAZING conversational norm, please adopt
The version of this rule I’ve adopted is “ask 2 clarifying questions before contradicting” (for high-level claims, not pure factual claims). This gets most benefits of both. And if I already know someone won’t have a useful clarification I shouldn’t be arguing with them anyway.