Nothing says "using the power of AI to create a reliable source of unbiased knowledge" quite like having to tweak your creation continually so that it won't accidentally veer into liberalism when the real-world facts steer the model in the "wrong" direction.