Exactly. These models don’t hold coherent views. You can prime any of them to agree with any view.
Which is good as it is bad.
Me: "Please make an app that does X in C"
LLM: "C sucks donkey balls, use Rust instead".
It's hard to have a general purpose tool that both has and does not have opinions.
Which is good as it is bad.
Me: "Please make an app that does X in C"
LLM: "C sucks donkey balls, use Rust instead".
It's hard to have a general purpose tool that both has and does not have opinions.