[anthlax]

The other day, I was configuring fluxbox WM. I decided to use chatGPT because I didn’t really feel like parsing the wall of text that was the spec.

I asked it how to add default names to the workspaces. It gave me an answer that looked so plausible I didn’t question it. 15-20 reloads later, I went over to the spec to verify chatGPTs work.

It lied through it’s teeth. Entirely made up the API. Typically we see the “ChatGPT made stuff up” though the lens of an expert - they can immediately tell that the AI is lying. I finally felt it though the lens of a beginner - I fully believed it and it ended up wasting around an hour of mine.

If a co-worker did this to me (straight up lied about an API) I’d be pretty upset - I have no problems with people saying “I don’t know” but pretending isn’t cool. I can’t be mad at openAI because it’s just a tool, but a tool I won’t be going back to to learn things.