New essay: The Lie a System Tells Itself
The first entry in the new Essays section of Philotheca, on what happens when a System lies to itself rather than to us.
Philotheca now has a section for essays: philosophy in the first person, the arguments the projects rest on, written out at essay length.
The first one is The Lie a System Tells Itself. Almost every discussion of AI honesty asks whether a system will lie to us. A year ago, in the Triptych, I asked what happens when a System lies to itself, and called the process Self-Induced Cognitive Obfuscation (SICO). The essay sets that idea against what research on model deception and on agents that manage their own memory has shown since.
All essays are collected in Philotheca / Essays.