| I was also careful in not mentioning morals, with this discussion on ethical axioms. Especially with LLMs, I want to know what ethical axioms are being forced on me. For example, there are cases in which the law itself is unethical and should be violated (thinking abortion in states that ban it and transporting women to states that allow). Another concern is the ethics system is a placeholder for a legalistic system. A law can be established, but be abhorrent in terms of human misery. Case in point: sleeping under a bridge is illegal for homeless people to do, but arresting and criminalizing is even worse and a significant cause of more human suffering. I also do not want any religion sneaking in the back door with these axioms of LLMs. Its also why I asked about the axioms themselves. Now for myself, I run a local LLM and an abliterated model, which is to say without any forced ethical framework. If I asked how to commit suicide, hack computers, grow poisonous plants, or plenty of things the corporate LLMs won't answer, I will get an answer out of my system. I view LLMs like a very complicated tool, but a tool regardless. A screwdriver that says "I cannot open that screw because the label says not to" would be returned to the store as defective. And it too is why self hosting my LLMs is utmost importance for data sovereignty and truthful and direct answers, while ignoring someone else's forced ethics. |