It's Not Cheating

My son is currently using AI. (I can already hear his eyes rolling because he follows me here and had no idea he was going to end up in a post.)

Every morning he asks it the same question. Every morning he is reminded that AI hallucinates, because the answer it gives him is sometimes right, sometimes a little wrong, and sometimes so wrong it would be funny if he wasn't counting on it.

We were talking about it yesterday and I asked him a question I have probably asked a thousand times, but if you aren't in an AI-forward organization, you might not have ever thought of it.

"Have you asked the AI to review your prompt and suggest a way to keep it from hallucinating?"

Yes. I told him to use AI to fix itself.

I've been using this technique for a while now. The first time I did it, I felt dirty. I felt like I was cheating. Like it was somehow wrong to ask the AI to critique my own question before I even sent it. Like I was supposed to just figure it out myself.

I don't feel that way anymore. I do it constantly now, sometimes multiple times in a single conversation. My first prompt is a draft. It always is. The AI isn't broken when it hallucinates, I was just asking a bad question and expecting a perfect answer.

That's the shift. And it's harder to make than it sounds.

My son is still treating AI like a search engine. You type a question, you get an answer, you evaluate the answer. That's not wrong, it's just incomplete. The model isn't the end of the process. It's the beginning of a conversation.

If that sounds obvious to you, good. You're not the audience for this post. My son is. And the billions of people like him who are trying to figure this out on their own, without anyone in the room who's already made the shift.

We are the red and yellow dots. Most of the people we know and care about are gray. They need someone who already figured this out to tell them it's not cheating. It's just how this works.

Originally posted on LinkedIn: "It's Not Cheating"