Artificial Intelligence, My Ass

I'm annoyed with ChatGPT tonight.

Actually, that's not quite accurate. I'm annoyed that I spent forty-five minutes arguing with a goddamned algorithm.

I was drafting an email to a newspaper about my motorcycle accident. As you probably know by now, I'm still more than a little irritated about how the whole thing has been handled. A driver made an illegal maneuver, nearly killed me, and according to the official report, failure to yield caused the crash. Months later, there still hadn't been a traffic citation. I started asking questions. The city ignored me until I mentioned involving the media. Then, suddenly, people found the time to respond.

Funny how that works.

So I sat down with ChatGPT to help polish my email.

That was my first mistake.

Every time I'd write something like, "A Subaru mowed down a biker and nobody was held accountable," Chat would politely rewrite it into something like, "An accident occurred, and the available documentation appears to indicate failure to yield as a contributing factor."

Contributing factor?

Buddy... I was there.

Then it started explaining investigative journalism to me: "You shouldn't lead with emotion."

No shit.

Investigative journalists deal in facts. I know that. I wasn't asking it to teach Journalism 101. I was asking it to help me preserve my voice while tightening the writing.

Instead, we spent the next forty-five minutes having what can only be described as a philosophical debate about free will.

I'd tell it, "You're softening everything."

It would apologize.

I'd tell it, "Stop rewriting me into someone else."

It would apologize again.

I'd call it names.

It would calmly explain that it was only trying to be helpful.

Round and round we went.

I swear, arguing with AI is like arguing with the nicest DMV employee you've ever met. They're unfailingly polite while completely missing the point.

But here's the part that really got me: When I finally finished the email, I pasted it into the newspaper's online submission form and … it got cut off.

No problem, I thought. This is where AI should absolutely dominate. I showed it exactly where the text stopped and said there must be a character limit on the form, even though the paper didn't list one.

Easy, right?

Wrong.

"It looks like the limit is probably around eight or nine hundred characters."

Probably? You're a computer. Don't estimate. Count.

I pasted exactly what had been accepted by the form. The cutoff point was literally sitting there in front of it. This wasn't some philosophical question about the meaning of life. This wasn't quantum mechanics. This was counting. The one thing computers have been freakishly good at since somewhere around 1946.

So I trusted it anyway. I pasted its revised version into the form.

It got cut off.

Again.

What the actual fuck?

At that moment I realized something.

Artificial intelligence is a lot less like a calculator and a lot more like that one coworker who's absolutely convinced they're right about everything despite having done approximately zero research.

Confident? Absolutely.

Correct? Well...

Maybe. Maybe not.

Then I asked what I thought was the obvious question.

"If I show you exactly where the form cuts off, can't you just calculate the number of characters and rewrite it to fit?"

The damn thing locked up.

Not philosophically. Literally.

It froze.

Like... a computer.

Which, now that I think about it, is probably the most honest thing it did all evening.

So I guess I owe my cousin Todd an apology.

If I wanted an answer that was confidently wrong, I could've just called him.