Reaction
Deception and loss of monitorability in AI safety
In a reply, a commenter suggests discussing deception alongside loss of monitorability and recommends a paper on monitoring.
TLDR
A commenter asks where monitoring will be discussed and suggests deception and loss of monitorability as a fit. They link to the paper “Monitoring Monitorability” and recommend Boaz Barak’s AI safety course as another resource.
Combined views
138
1 Source, first seen ago
