Goody-2 is the world’s most responsible AI model. It’s a parody showing how ridiculous you can get if you try to ensure “AI Safety”. The real world can be messy and dangerous. Real people say things that other people dislike. But once you decide you need “protection” from the bad stuff, the only question is who you trust to do the “protecting”. This LLM may seem outrageously protective, but the kernel of truth in the parody is to show what road you’re headed when you veer into the direction of protection over truth.

We all want our AI conversations to be helpful. Nobody wants a conversation partner who is rude, abusive, or misleading. In the real world, we simply ignore those whose speech we don’t like. Shouldn’t it work that way in the LLM world?