AIRoutine office work versus alleged wrongdoing in AI training data
One post suggests that records of alleged wrongdoing may outweigh routine workplace material in AI training data, potentially skewing its picture of business communication.
TLDR
A post asks whether AI training data contains more ordinary workplace records—meetings, summaries and action items—or emails, decks and memos presented as evidence when people in companies are accused of wrongdoing. The author suspects the latter dominates and tilts the view of business communication. They draw parallels with newspaper archives and medical case studies, arguing that these emphasize exceptional events and unexpected outcomes over routine ones.