Let Them Eat AI

How my demonstration of why not to use ChatGPT backfired

A couple of months ago, when mainstream media finally decided that AI was a “threat,” CNN called to interview me about the typical, first-order fears: kids cheating on papers, the loss of jobs, impact on the economy, and so on. I took the opportunity to do that thing where you ask ChatGPT to answer the question for you.

So while Jake Tapper’s correspondent asked me his first question, I typed it into ChatGPT and read the result. Of course, after we went back and forth a bit about the inadequacy of that answer, I revealed that it had come from ChatGPT.

My point was that it wasn’t the stock answer that an audience wants to hear, but the human opinion. Or, at least that’s how we should be thinking about these questions. Because only then can we challenge the underlying assumptions on which our questions rest. ChatGPT only looks at the past. It generates the most probable word response to a question based on an “average” of everything already said on the issue. It’s pure reversion to the mean.

Only a human has the ability to challenge all that history. So when CNN asks “what about the unemployment problem?” I can give them the typical AI answer, but then respond as only a human can: “What if unemployment is a solution? I’m fine for computers to do…