Apparently a historian at a college near my own has her students "interview" historical figures like Thomas Jefferson and whatnot by asking questions of LLMs that have been "instructed" to respond in the voice of those people. Jesus wept.
Apparently a historian at a college near my own has her students "interview" historical figures like Thomas Jefferson and whatnot by asking questions of LLMs that have been "instructed" to respond in the voice of those people. Jesus wept.
that makes me feel all sorts of weird--it feels like this kind of nonconsensual digital resurrection rife with ethical issues.
It assumes that the LLM can somehow know the totality of a person's thoughts through the portion of their written words that have survived -- and of course, what's the guarantee that the body of texts hasn't been chosen to favor particular narratives / interpretations?
The extremely deterministic methods of LLMs -- basically a probability calculator for words -- also seem to fly in the face of the turn toward "contingency" as a historical explanation in the last few decades.
But also, what of marginalized, and/or racialized historical subjects? I dread to think how they might end up being represented. Why are we being so credulous and uncritcal about this all, sigh
Did you see this piece of credulous (sorry to borrow your term) fluff, from an honest-to-goodness historian? www.newyorker.com/culture/the-...
ugh, no. Will have to look at it. I will only say that I will need professors at Ivies to understand how AI will disproportionately devastate the humanities at non-Ivies.