I agree. This is just make believe based on a smaller subset of human writing than LLMs we have today. It's responses are in no way useful because it is a machine mimicking a subset of published works that survived to be digitized. In that sense the "opinions" and "beliefs" are just an averaging of a subset of a subset of humanity pre 1913. I see no value in this to historians. It is really more of a parlor trick, a seance masquerading as science.