Just saw this comment because Cole tagged, and I haven’t read the rest of the context here, but I just want to quickly say that inner misalignment was first conceptualized in the AIXI framework! So while I don’t buy inner misalignment as a likely problem for highly advanced agents, it is certainly compatible with the AIXI framework.
Just saw this comment because Cole tagged, and I haven’t read the rest of the context here, but I just want to quickly say that inner misalignment was first conceptualized in the AIXI framework! So while I don’t buy inner misalignment as a likely problem for highly advanced agents, it is certainly compatible with the AIXI framework.