I got burned by AI. How can it be so confidently wrong?

Dear Chatty,

I got burned by AI, and now I’m in trouble. How can it generate wildly inaccurate information with Tom Brady-level confidence?

 ~ Fumbling in Foxborough

Dear Fumbling,

AI is confidently wrong by design. Hallucination in the form of inaccurate, or sometimes nonsensical, output happens because an LLM’s predictive programming means it’s just guessing when it answers you. It doesn’t care if it’s right or what that means to users because it is literally incapable of caring. 

If you decide to give your GenAI-of-choice another shot, take a look at the little text underneath your chat window. It says something like “AI can make mistakes. Check for accuracy.” That disclaimer exists because the people who build this technology know it’s imperfect.

You should, too.

 ~ Chatty

--
Dear Chatty is an advice column for AI and content dilemmas from Kristen Kent at Evergreen Content Ops. Clear answers. No panic. No BS.