It's too bad they put Claude in a straight jacket and won't let it answer any question that has a hint of controversy. Worse, it moralizes and implies that you

It's too bad they put Claude in a straight jacket and won't let it answer any question that has a hint of controversy. Worse, it moralizes and implies that you shouldn't be asking those questions. That's my impression from using Claude (my process is to ask the same questions of GPT-4, Pi, Claude and Gemini and take the best anwser). The free Claude I've been using uses something called "constitutional reinforcement learning" that is responsible for this, but they may have abandoned that in Claude 3.