Is AI unreliable?

— Is "human" always better?

Spoiler: no.

“AI hallucinates.”
“AI produces a ton of slop.”
“Everything you do with AI is flat and repetitive.”

Ok, there could be something true in all these sentences. AI work is not always perfect, it has to be checked against reliable sources, and it should be - at least - re-read and maybe edited a bit before publishing anything. Be it a text, an e-mail, or a piece of code, you should always validate it, because it’s the only way to be sure not to publish crap.
But this is not a new problem at all.

In my 20+ year career I worked with many people, and most of them had the same identical problem: they invented things to cover their ignorance, unable to admit they didn’t know something. When you know very little, you can’t spend your days saying “I don’t know”, so you stop admitting that and start pretending to be as good as your interlocutor. To be able to admit you don’t know something, you have to be conscious that your value won’t decrease by that affirmation, because people in the room know how much you know.
Now we call it “hallucination”, but we had to do with it for years, when it was only “pretending to be smarter”.

And we can’t say that all the human content is good: the internet is a latrine since it was invented, containing all kinds of human idiocy and giving all the people the same space to voice their “brilliant” ideas. It’s democracy, babe.
So we had to understand how to filter this flood of trash, even before AI was available to the general public. Not all the people (euphemistic) are able to apply such a high level of judgment, so the problem is real: but it’s not about the AI, it’s about the human mind, that’s not always as developed as we like to think.

Also when the content is not crap, we can’t say that it’s always brilliant: it’s not a binary choice.
A couple of days ago my YouTube feed showed me a video of a middle-aged “marketing guy” ranting about AI and about the fact it produces flat and conformist graphic design. What he was forgetting is that AI is trained on human content, and if it produces conformed graphic design, it’s because it always saw the same things during its training.
Human content is flat and conformed too, simply because there are far more average graphic designers than brilliant ones (or journalists, or coders, say yours). But if a human gets paid for crappy work, it’s good: if an AI gets paid a fraction of the cost to produce the same flat result, in a fraction of the time, it’s not good anymore, because that average human being, who is doing nothing to elevate herself, should continue to produce her things and be paid for that.

The problem is not about humans vs AI: the problem is that when we publish something, we have to be proud of what we are giving to the world.
It’s not a thing if I write this article by hand - and I’m doing so, because I like to do it - or if I take my ideas and ask the AI to investigate them with me to write the article, or if I ask the AI to check my English. The thing is that I have something to say; if I haven’t, it’s really better to save bytes at all.