Reaction
An argument for prevention over after-the-fact AI alignment fixes
A post claims 99% of current alignment work is after-the-fact treatment and questions whether that approach is sustainable.
TLDR
A user argues that 99% of current alignment work treats problems after the fact rather than preventing them, and warns against assuming that approach will be sustainable. The post calls for a longer planning horizon and asks what it takes to “raise a good person” as a way to think about alignment’s roots.
Combined views
—
1 Source, first seen ago
