The developers are going broad rather than deep, firehosing a bunch of varied historical figures into their library, which is probably a smart commercial decision - as there is broader appeal, something for everyone - and maybe the idea is to attract more “gee whiz” novelty subscribers to try it than to retain loyal & engaged users, but even after my first use, I hit a wall of - no matter who I tested - repeated verbatim boilerplate answers to very different questions, that weren’t even historical quotes but still AI contrived. At best, you’d think the AI could act as a super fast/smart librarian and reference an archive of historical accounts & direct firsthand comments from this person in many cases (surviving memoirs, diaries, letters, scriptures attributed to them in some cases, etc). I think the trouble is that history is too edgy for the current “safeguards” (censorship) put on AI & everyone’s super concerned with being held accountable for what AI says even in character. Which always makes it worse than the lamest of high school plays, because it’s super modernized, sanitized and dishonestly “safe” per our current ethics & basic world view. Some issues aren’t even ethical, they’re just not reflections of modern scientific consensus and it simply refuses to discuss them - like ghosts, for instance. John Dee could not elaborate on any of the entities he communicated with or how. If the goal is accuracy, let it quote sources based on a photographic memory of all archived data on John Dee. If, on the other hand, we wanted it to be *less* accurate and more “fun”, let it synthesize responses in the dialect & cultural attitudes of that period, based on the historic data we have. But the training data on all the figures I tested was lamentably shallow, both in terms of the cultural/socioeconomic/environmental context of their eras and indeed of their personal views and interests in life. It’s more like a badly structured, poorly edited history book lite “for kids!” than anything of meaningful substance. Disappointing bland commercial misuse of AI potential, as usual.
Show more