Comment by nojvek
2 years ago
Claude refuses a lot. GPT4 also refuses a lot and one has to try several prompts to get out what you need.
LLMs are trained on the entire internet and more.
I want a model that just gives me the answer with whatever it knows instead of playing pseudoethics.
Sure it can say this is dangerous “don’t do this at home” but let me be the judge of it.
But aren't you a small child, and doesn't the AI know so much more than you?
To be honest, what they view as ethical is actually unethical: this idea that the AI knows more than a human, in the human's situation, and can pass judgment on that human.