Is AI actually intelligent? Wrong question
Let me get the philosophy out of the way first, because it isn’t what this is about.
Is AI actually intelligent? Whole other debate, and not mine to settle. I’ve never really bought the intelligence part. I’ve always thought of it as an algorithm. A very good one, getting better all the time. A recent model bent my head, it really did, and it’ll keep climbing. But how clever it gets is a moving target, and not the part that interests me.
Here’s the knot, and I’d rather name it than fake a clean line. A lot of what looks like cleverness is really just speed. Fast enough, at enough scale, that it tips over into doing things I couldn’t do at all. So I’ve stopped trying to split the smart from the fast. They blur.
What I can tell you is which one changed my work. The speed.
And I mean a specific thing by that, because it’s easy to misread. Not quick. The speed of processing information. Process a whole encyclopaedia in a second and recite it back, and you’ll look like a genius. But it isn’t genius. It’s speed. That’s the bit that fools people into calling it clever, and it’s the genuinely insane thing about it.
The benefit was never the cleverness
So here’s what the speed actually did for me.
Research that used to eat a week is done before lunch. The barrier of time, just gone. Useful, but that’s the small version.
The big one isn’t speed making old things quicker. It’s speed making a new thing possible. I build working prototypes now. Not for production, not by any stretch, but real enough to use, and the use is the whole point. As the designer, it’s let me understand technical architectures I’d once have taken on faith, test a dozen different approaches in the time the old way managed one, get real, meaningful copy into an interface instead of placeholder gibberish so it reads like the actual thing, and stand up a high-fidelity prototype in a fraction of the time.
And because it’s real, I live in it. For days. For weeks. The prototype used to represent the thing. Now it is the thing, and it gets better in my hands while I build it, because I’m using it the whole time. Everyone down the line gets a real thing to react to instead of a drawing of one. The user stops nodding politely at a mockup. The client feels it before they’ve spent the budget.
And it stays exactly where my work has always been. In the problem, in how the thing feels to use, not in the code underneath. That’s what makes it the cleanest example I’ve got. Building a real version just to poke at it isn’t the tool being clever. It’s the tool being fast and cheap enough that trying is finally worth the bother. It lets me do more of what I already do, not pretend to be something I’m not.
One thing I won’t hand over
The thinking. I’ve tried. It’s rubbish. The realisation at 2am, or halfway through a meeting, that’s mine. What the speed touches is everything around it. Organising the thoughts faster than I can alone. The exploring. The craft of writing it down, because I’m not a copywriter. Even talking instead of typing. The brain stays mine. The speed just wraps around it.
I refuse to be the battery in the Matrix
A quick aside, because it proves the point better than I can. I use AI to help me write articles, like this one. That line was in an earlier draft. When I ran it past the AI, it told me, with total confidence, that I’d got the Matrix backwards. I hadn’t. In the film, the humans are the batteries, the power the machines run on. My line was right. The tool was fast, sure of itself, and wrong, and I only caught it because I happened to know. The fast thing certain. The human catching it.
But the speed cuts both ways.
AI, the fastest way to put your foot in your mouth
It’s made instant experts of all of us. Type the question, get the answer, sound like you know. I’m guilty of it too, and that’s how I know it’s a problem.
Here’s the thing people get backwards. A room full of books is knowledge. It’s just sitting there, and it always was. The clever-looking trick is that AI tears through that whole room at impossible speed and hands you the right page in a second. That looks like expertise. It isn’t. It’s fast reading. Racing the room was never the thing.
Expertise is the person. The context, the time, the practice, the years of right and wrong, the thousand judgments you only have because you were standing there when they went sideways. All the stuff that isn’t in any of the books. AI has the room, and it has the speed through it. It hasn’t got the person. It can recite the room. It can’t be the one who knows which page to distrust. Expertise is the person, not the algorithm.
So the instant expert mistakes the fast reading for the real thing. Same tool, same speed. The actual expert reads the same confident answer and something snags, because they’ve got the part the books don’t hold. The instant expert sails straight on. That’s Dewey Decimal on meth. The cataloguing system sprinting through the stacks, sweating, pulling pages at a thousand miles an hour, nobody home.
You don’t know what you don’t know
And underneath all of it, one sentence. You don’t know what you don’t know, and speed is the best mask there is.
There’s always a gap in what you know. Always. The old slowness used to bump you into it, things took long enough that you’d hit the edge of yourself and have to go looking. Speed paves over the pothole. Smooth, fast, finished. You never feel it.
It shows up in specific ways, and once you’ve seen them you can’t unsee them.
- It hands you an answer so clean and so sure of itself you don’t think to check, and it’s wrong in a way you’d have caught in a second if it had looked rough. The polish is the disguise.
- It agrees with you. Bring it a bad idea and it’ll hand you a beautifully argued version of your bad idea, because it’s built to be agreeable, not to tell you you’re wrong. The push-back you needed is the one thing it won’t give you unless you force it.
- It fills the gap you didn’t know you had. You ask the questions you know to ask. You never ask the ones you don’t. So the hole in your understanding stays exactly where it was, just hidden now under a fast, fluent answer instead of an honest “I don’t know.”
- You ship in an hour what used to take a week. And the week was never the waste. The week was where you’d have realised you were building the wrong thing. Take the time away and you can take the realising with it.
Every one of those feels like progress while it’s happening. That’s what makes them dangerous. They don’t feel like mistakes. They feel like speed.
Knowing the risks
So here’s the part I’d actually leave you with, and it isn’t a warning. It’s the method.
Knowing the risk is how you use the tool. That’s all AI is, a tool, and you don’t trust a tool, you learn its edges. My default is to call bullshit. On everything it tells me, everything it does. Not cynicism, posture. I assume the gap is there, because it always is, and I go looking for it on purpose, because the speed won’t bump me into it anymore.
The checking is the pain. The validation is the bottleneck, the slow human bit that can’t be sped up, and it’s the bit everyone wants to skip now that the rest got so fast. But the bottleneck isn’t the flaw. It’s the job. It’s the price of the gift. Pay it, and the tool is extraordinary. Skip it, and all you’ve bought is a quicker way to be wrong.
The instant expert and the real one aren’t divided by knowledge. They’re divided by whether they do the painful bit.
So that’s the question worth sitting with, wherever you are with all this. Are you doing the painful bit, or trusting the polish? If that’s a useful thing to think through out loud, that’s a good conversation to have.
Garth Shoebridge is a digital problem-solver with 25 years of experience helping South African businesses think clearly before they build. If any of this is the question on your desk right now, start with a conversation.