I mean I won't disagree with you but I think a more fundamental issue is that we are so easy to lie to. I'm not sure it matters whether the liar is an AI, a politician, a corporation, or a journalist. Five years ago it was a bunch of people in office buildings posting lies on social media. Now it will be AI.

In a way, AI could make lie detection easier by parsing posting history for contradictions and fabrications in a way humans could never do on their own. But whether they are useful/used for that purpose is another question. I think AI will be very useful for processing and summarizing vast quantities of information in ways other than statistical analysis.

AI is only as good as the model it is trained on, so while there are absolute truths, like most scientific constants, there are also relative truths, like “the earth is round” (technically it’s irregularly shaped ellipsoid, not “round”), but the most dangerous “truth” is the Mandela effect, which would likely enter the AI’s training model due to human error.

So while an AI bot would be powerful, depending on the how tricky it is to create training data, it could end up being very wrong.