Nvidia has trained an LLM on their codebase to save senior designers having to answer junior's questions about what some code does.
I don't know how many times i have to say this, but these are not search engines. They are not inference engines. Their answers are 100% "what would be a plausible sounding string of words to respond to that input?" that's it. That's all they do.
