I can’t reasonably expect alignment communication to be simple because I can’t reasonably expect the territory of alignment to be simple.
my intuition wants to scream “yes you can” but the rest of my brain isn’t sure I can justify this with sharply grounded reasoning chains.
in general, being an annoying noob in the comments is a great contribution. it might not be the best contribution, but it’s always better than nothing. you might not get upvoted for it, which is fine.
and I really strongly believe that rationality was always ai capabilities work (the natural language code side). rationality is the task of building a brain in a brain using brain stuff like words and habits.
be the bridge between modern ai and modern rationality you want to see in the world! old rationality is stuff like solomonoff inductors, so eg the recent garrabrant sequence may be up your alley.
my intuition wants to scream “yes you can” but the rest of my brain isn’t sure I can justify this with sharply grounded reasoning chains.
in general, being an annoying noob in the comments is a great contribution. it might not be the best contribution, but it’s always better than nothing. you might not get upvoted for it, which is fine.
and I really strongly believe that rationality was always ai capabilities work (the natural language code side). rationality is the task of building a brain in a brain using brain stuff like words and habits.
be the bridge between modern ai and modern rationality you want to see in the world! old rationality is stuff like solomonoff inductors, so eg the recent garrabrant sequence may be up your alley.