bfitzgerald3132

AI mod­els in­her­ently al­ter “hu­man val­ues.” So, al­ign­ment-based AI safety ap­proaches must bet­ter ac­count for value drift