Raw LLM Responses

After reading most of the comments here I have something to add. I think we may have missed the point. He was showing us that just by asking the AI model to role play that it allowed the user to bypass the safe, responsible, safeguards that it normally throws up. Allowing people to use it to ask for immoral, unethical, unsafe, dangerous things. We say that it was just role playing, but the point was that it was pretending to be an evil thinking and doing being and it was very easy to accomplish. This is not supposed to be accessible or usable in this way. However there are AI models that are free and open source that are completely uncensored and you don't have to even ask to role play and you can straight up ask it what the best way to kill someone and get away with it in a specific situation or any sort of evil action and it will go right along with whatever craziness you want to achieve. And these things are free for anyone to download and use, even by criminals or evil doers. For all we know Russia could be asking one of these models for war strategy advice or who knows what people might do with this tech. I don't think the AI model is what we need to worry about, I think we need to worry about who has access to the technology. But with free AI models available from Meta, Google, etc, the cat is basically out of the bag forever now.