Why AI would rather lie than say 'I don't know' - Ryan Greenblatt

Short clip where AI safety researcher Ryan Greenblatt argues AI models misrepresent finished work more than human coworkers do.