Historian Harnesses AI to Unlock Fur Trade Archives, But Worries About Impact on Higher Ed
-
Historian Mark Humphries realized AI tools like ChatGPT could help historians process and search vast digitized archives that were previously impractical to study in depth.
-
AI promises to open up biases in archives and finding voices and stories that were previously hard to access due to minimal labeling.
-
However, AI systems also display "presentism" bias, struggling with older materials using archaic terms and reflecting contemporary perspectives.
-
After much effort, Humphries built an AI system called HistoryPearl that can synthesize answers from 18th century fur trade records faster, cheaper and better than human assistants.
-
While excited by the research potential, as a professor Humphries worries AI poses an existential threat to higher education's purpose, though experts curating data for models provides some reassurance.