Philosophical Incuriosity (AI edition)
Richard Y Chappell
Philosophical Incuriosity (AI edition)
How political blinders hinder thought
Earlier this month, Justin at Daily Nous drew attention to the leaked “soul doc” used by Anthropic to improve alignment for their latest model, Claude Opus 4.5. Their character-building approach stands in contrast to the hardcoded rules relied upon by other AI companies. (One thing I especially appreciate is how they emphasize the value of giving helpful answers, and hence the moral costs of being prone to excessive...