This is not a newsletter about ChatGPT (if you want to read one of those, I’d recommend this one), but our topic this time around was prompted by reports of the chatbot just plain making up “facts.” That has disturbing implications—if there’s one thing we’ve already got plenty of, it’s online misinformation—but it didn’t surprise me. After all, for nearly two decades, I’ve been bombarded with confident-sounding nonsense spewed by the translation apps that supposedly threatened to render my work obsolete.
Credit where credit’s due: Google Translate and similar tools have improved somewhat since the ubiquitous “paper jam” / “mermelada de papel” screenshot was taken. But being programmed to recognize commonly used two-word phrases is not the same thing as understanding context, much less recognizing nuances or errors in a source text.
Continue reading this post on Substack. (Don’t worry—nothing is behind a paywall!)