@bayindirh @jenniferplusplus

I think I do understand.

I have to use #LLMs at work, because (a) they're the front end to some of the systems I have to use to do my job, and (b) I'm assessed on whether I use "AI". Obviously, LLMs have no understanding, no mental model of the problem or its solution, no consciousness and no reasoning ability, and they get things wrong depressingly often. But possibly because they're so fallible, I can see the potential for addiction. Just like the gambling addict pouring a week's wages into a slot machine that keeps nearly paying out the jackpot — one more spin, bro! — it's tempting to keep trying one more prompt in the hope of hitting gold. And it's hard to remember how unreliable these machines are when every answer is delivered with reassuring confidence and a gleaming smile.

I respond by using LLMs as little as possible and never outsourcing my thinking by, for example, letting them design code or write a document.

An LLM may feel like your friend, but it isn't. An LLM is to friendship what icing sugar is to nourishment.