"Did you train the model using real world data?"

"Yes. We want the model to be useful in real world applications."

"Then it is biased. The model is biased because data it was trained on was generated by people and people are biased. There is no such thing as an 'objective' model, just a model that is biased in a different way."

The left-wing bias of ChatGPT probably derives from intentional RLHF, not from the training text. If RLHF raters are woke, the fine-tuned model will be too.

> RightWingGPT was designed specifically to favor socially conservative viewpoints

Thats how these things work. You steer them.

> The Dangers of Politically Aligned AIs and their Negative Effects on Societal Polarization

We are still really talking about the dangers of humanity.