Harvard, MIT: AI's Potemkin Understanding

very nice discussion of an intriguing paper covering some of the limits of the current llms.