The Idiot’s Guide to Using AI

Artificial intelligence may be one of the most sophisticated technologies humanity has ever created, which naturally means we immediately found a way to use it badly.

This should surprise absolutely no one.

We invented smartphones and used them to photograph our lunch. We gained access to practically the entire accumulated knowledge of civilization and somehow turned the internet into a place where strangers argue about whether the Earth is flat. Now we have artificial intelligence—technology capable of analyzing enormous amounts of information, helping people write, research, brainstorm, calculate, organize, code, translate, and solve problems—and some people are using it like a Magic 8 Ball with Wi-Fi.

“AI, tell me what to do with my life.”

Well, Kevin, perhaps we could narrow that down a little.

The problem is that people have been sold the idea that AI is supposed to know everything. Ask question. Receive wisdom. Done.

Except AI doesn’t work that way. It is enormously capable, but capability is not clairvoyance. If you give it a stupidly vague instruction, you should not be surprised when it gives you a beautifully formatted stupidly vague answer.

“Write me something good.”

About what?

“You know.”

No. It fucking doesn’t. This is perhaps the first great rule of artificial intelligence: AI cannot retrieve information you have chosen to keep inside your skull.

Then there are people who treat the first answer AI produces as though Moses just came down from Mount Sinai carrying it on stone tablets.

AI says something. They copy it. Paste it. Send it.

No fact-checking. No editing. No momentary consideration of whether an actual human being would ever write, “I hope this message finds you well. In today’s rapidly evolving landscape…”

Nothing announces I outsourced my personality quite like suddenly sounding like the regional vice president of Synergistic Solutions Incorporated.

AI gave you a draft, Brenda. It did not give you a personality transplant.

And that may be the strangest part of all this. People complain that AI writing sounds like AI while making absolutely no effort to tell AI how they themselves sound.

If I hired a ghostwriter and told him, “Write my autobiography,” then disappeared for six months, I would have very little standing to return and complain that the book doesn’t sound like me.

What did you tell him about yourself? Nothing.

Did you show him your writing? No.

Did you explain your humor? No.

Did you edit anything? No.

Then congratulations. He wrote his autobiography. Your name just happens to be on the cover.

Using AI well requires something deeply inconvenient: you still have to think.

I know. Horrifying.

You have to provide context. You have to ask better questions. You have to recognize when an answer smells suspicious. You have to say, “No, that’s not what I meant.” You have to revise, challenge, redirect, and occasionally tell the machine that what it just wrote sounds like a motivational poster hanging inside a dentist’s office.

That is not AI failing. That is you using the tool.

A calculator does not eliminate the need to understand numbers. GPS does not mean you should drive directly into a lake because the pleasant electronic lady said turn right. Spellcheck has existed for decades and humanity still cannot decide which “there” belongs in a sentence.

Tools amplify the person operating them. And there lies the uncomfortable little problem.

Give a smart person AI and you may get better research, sharper questions, faster work, and ideas that would have taken days to develop.

Give an idiot AI and now you have an idiot operating at broadband speed.

That is progress, technically.

The dangerous idiot is not even the one who doesn't know something. We all don't know things. The dangerous idiot is the one who no longer realizes that not knowing something matters because ChatGPT produced four paragraphs about it in twelve seconds.

Suddenly he is an epidemiologist. Tomorrow, constitutional scholar. Friday, geopolitical strategist. Saturday morning, nutritionist. By Sunday he has diagnosed the strange noise coming from your transmission and would like to discuss quantum mechanics.

Confidence has always been cheap. AI merely introduced same-day delivery.

None of this means AI is useless or that using it somehow makes a person lazy. That argument is equally stupid. Human beings have spent our entire existence inventing things specifically so we can do less work. The wheel was basically an early productivity hack. Nobody looked at the first wheelbarrow and shouted, “CHEATER! Carry those rocks with your bare hands like your ancestors!”

The point of a tool is to make you more capable. But outsourcing labor is different from outsourcing judgment.

That distinction matters.

AI can help you write the email. You still need to know whether you should send it. It can summarize the contract. You still need to understand what you're signing. It can challenge your argument. You still have to decide whether your argument survives. It can give you twenty ideas. You still have to recognize which three aren't bullshit.

But somewhere in the process, a human being still has to look at the thing and say:

“Okay, that part stays.”

“That part sounds like ChatGPT.”

“Delete that shit.”

“Too preachy.”

“Make it funnier.”

“There. Now it sounds like me.”

That isn't cheating. That's editing. That's judgment. That's using the damn tool instead of allowing the damn tool to use you.

Artificial intelligence was never going to eliminate human stupidity. That was an absurdly ambitious feature request.

If anything, AI has simply given stupidity better grammar.

And God help us, now it has bullet points.