Truthful AI

With respect, I would encourage some background reading and deeper thought around what you are describing as bias, particularly around human rights and agency. There are some "common" values but there are also sides. Either you believe all humans are endowed with inalienable rights, or you do not. Those are sides. The "value" of tolerance is not shared, so violent preclusion of rights will be applied by one side to the other. In that context, I would encourage reading the "ChatGPT is bullshit" paper and some of its references. "Bullshit" is defined precisely. The key point is that LLMs are compression algorithms for content they have been trained on. They can be engineered to generate debate material for one "side" or another based on their training material, but those debates are value-less conducted by "entities" without stake. If one LLM is producing material for the side that women are not humans with bodily autonomy and one is producing material for the side that they are- and you are following and deciding your view based on that "debate" you are deceiving yourself, because you are already on the side that women are not people entitled to rights. And the LLMs are not women, actual humans, with actual stakes. Within guardrails, LLM generated material can be useful but they cannot be entrusted with "debates." And claims of "truth" specifically should be viewed with extreme skepticism. Elon is without question on the side that not all humans have inalienable rights.