jwfiredragon
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
Three perspectives on corrigibility - shards, simulators, singular learning
According to the excerpt, jwfiredragon, in a post first drafted at the ML4Good bootcamp, asks whether a corrigible agent could be built. The author says it once looked impossible, but newer theories of how language models learn suggest it may be possible with care, and defines terms before comparing three perspectives.