That would just be a battle of biases to see which bias wins. I don't know that AI can reach any "truth". It's just statistically figuring out which word goes after another, at least in its current form.
Discussion
Right but perhaps that's the result. A descent into incomprehensible madness would illustrate the risk of relying on these things too much. Rather than a generative dialectic, you would likely get madness.