The One Filter
A single structural claim, that the resemblance of an output is not the identity of the process that made it, and that self-referential evaluation is …
xbard
Tagged: Alignment
xbard
Two ways to lose the upper regime
A new Frontiers in Human Neuroscience hypothesis paper proposes a biological mechanism for the switch between System 1 fast-parallel cognition and System 2 slow-sequential cognition. …
xbard
The Two Ends of Refusal
Two pieces on this site, written six weeks apart from opposite directions, turn out to be the same argument. What that argument actually is, and …
xbard
The Doss Protocol
A specification for consciousness preservation as a hard constraint on AI optimisation systems. Named for Desmond Doss. Working draft.
xbard