you continue to not engage with my points and bring up weird arguments An LLM is simply an interface to the knowledge (RAG) what? what does RAG have to do with this? an LLM is a tool for fast next-token prediction and as it turns out, next token prediction is a really useful proxy to compress a ton of information into a model that can output it very quickly. this has safety implications as part of this knowledge is harmful and these models can (and have) been used to act as agents. we have alrea...