I had AI book me a vaccine appointment. It found a pharmacy in our neighborhood and booked a time.
The pharmacy was in the neighborhood, technically. It was also kind of out of the way, and there was a closer one.
My wife thought this was hilarious. Her take, basically: this is why you can't trust AI.
The next day she booked her own vaccine and did the same thing. A pharmacy in the neighborhood, kind of out of the way, with a closer one right there.
I'm not going to pretend I didn't enjoy that a little.
It got me thinking. AI makes mistakes, obviously. But does it make more of them than people do?
A booking page shows you a list of stores "near you," and near covers a lot of ground. It's an easy one to get wrong. My AI got it wrong. One day later, so did a smart person who was paying attention.
When a person books the far pharmacy, it's a funny story at dinner. When AI does it, it's proof you can't trust the whole thing. Why are we so quick to jump on the AI one?
Here's what happened after my booking. I gave the AI some feedback, and it got a new rule: before it books any in-person appointment for me, it checks whether the location it picked is the closest one. If it isn't, it doesn't book. It asks me and gives me both options.
That was one fix, and it applies to every booking now.
A person's mistake gets laughed off and happens again. An AI's mistake can be fixed once, for good.
Also, I can give feedback to my AI without getting in trouble. My wife, not so much.
Benjamin Thomas
ForIT Team
