Is The AI Sure, Or Is It Guessing?

It never sounds unsure.

That’s the part that gets you. It answers everything in the same steady voice. The thing it knows cold and the thing it made up out of nothing come out sounding exactly alike.

You can’t hear the difference. That’s not your fault. There isn’t one to hear.

My mother got to where she couldn’t keep her pills straight. So I sat down at her kitchen table one evening with all of them lined up in front of me. Seven bottles. Three different doctors, none of them talking to the other two. Two of the bottles had the same thing in them under different names, and I only knew that because I’d been reading labels for an hour.

I typed the whole list into the AI and asked it to sort it out for me.

It gave me a beautiful answer. Clean list, times of day, little notes about which ones to take with food. It read like something a nurse had written out.

And a piece of it was wrong. Not the whole thing. A piece. Stated in the same calm voice as the rest of it.

I keep a list of the things a machine does that make it dangerous to lean on. This one sits at the top of it.

Here’s why it happens. The thing isn’t looking anything up. It’s finishing a sentence. It has read a mountain of writing and it has gotten very good at producing the next likely words. When it knows, it produces the right words. When it doesn’t know, it still produces words, because producing words is the only thing it does.

There is no little light that comes on inside it that says this part I’m sure of, this part I’m inventing. It doesn’t have one. So it hands you both in the same tone of voice and lets you sort it out.

And you can’t sort it out, because you came to it in the first place because you didn’t know.

Here’s the same question both ways.

Without a word from you, you paste in the list and ask it to sort your mother’s medications. Back comes a tidy schedule. Confident. Complete. You print it and stick it on her refrigerator. You have no idea which line on that page came from something real and which line it filled in because a line was expected there.

Now the other way. You paste in the same list. It tells you which ones it recognizes and can speak to. It tells you which two names it isn’t certain refer to the same thing. It tells you flat out that it cannot see her chart, doesn’t know her kidney function, and is not the one who should be answering the timing question. Then it tells you the exact question to walk into the pharmacy and ask.

Same list. Two different rooms. The second one is shorter and less impressive and it’s the one that’s worth something.

The fix is one line, and here it is free.

Before you ask it anything that matters, tell it this:

Tell me what you don’t know before you tell me what you do.

That’s the whole thing. It changes what comes back. Instead of a smooth page, you get a page with holes marked in it, and the holes are the most useful part, because now you know where to go ask a person.

I’ll say the rest of it plain, because this one’s not a productivity gripe.

That machine is not your mother’s pharmacist. It doesn’t get to be. Take the list to the pharmacy counter and hand it over — they’ll do it for free, it’s their actual job, and they can see things the machine can’t. Use the AI to figure out what to ask them. Don’t use it to skip them.

Now, the trouble with a line like the one above is that you have to remember it. Every time. And you won’t. You’ll remember it at the kitchen table when you’re paying attention, and forget it at eleven at night when you’re tired and worried and want somebody to just tell you the answer. That’s the night it matters.

That’s the honest weakness of one good habit. It only works when you’re already sharp.

So I put a short file together — a couple dozen of these written out plain. You hand it over once at the start of a conversation and then you just talk. Nothing to manage. It changes how the machine talks to you for the rest of the session.

I’ll be straight about what it does and doesn’t do. It changes the manner. It doesn’t make the machine right. Nothing makes the machine right. What a better manner buys you is a machine that shows you where it’s standing on nothing — and on a night like the one at that kitchen table, that’s the difference between a printed page and a real question.

It’s free. Two pages, no signup, nothing to buy.

You will see clearly now.

How-To-Use-The-Working-Manner-Layer

TFB-Working-Manner-Layer (3)Download

This post was drafted with AI governed assistance and reviewed and directed by Michael S. Faust Sr. before publication.

Post Library – Intelligent People Assume Nothing

Contact: micvicfaust@gmail.com