NOTES · Generative AI · Education · Technology

It's Artificial. It Isn't Intelligence.

Sergei Sikorskii · 19 June 2026 · 3 min read

“The real question is not whether machines think but whether men do.” — B. F. Skinner, 1969

One of my students recently asked whether I’m afraid that AI will take people’s jobs, and whether I, as a teacher, worry about becoming unnecessary. I’m not, though the reason isn’t that teaching is somehow safe.

We have lived through this before. Cities once employed lamplighters to light the street lamps each evening and put them out at dawn. There were knocker-uppers who went door to door waking workers for their shifts, switchboard operators who connected every call, typists, and plenty of other trades that simply faded out. But the people didn’t fade out. What disappears, when technology shifts, is a particular way of doing the work, and it never disappears in just one way. Some jobs vanish outright. Some are transformed into something different. And some simply pick up the new tools and end up faster, sharper, better at what they already did. So “will AI take our jobs?” has always struck me as the wrong question, or at least a lazy one. The more interesting question is what this technology actually is.

Is it really intelligence?

Here I hold a less popular view. I’m not convinced we should be calling today’s large language models “artificial intelligence”. Is it artificial? Definitely yes. Is it intelligence? Nope. Look at what actually happens under the hood and you find prediction: the system estimates the next token, the next word, the most probable continuation. We have seen this before in miniature. T9 on our old phones. Then autocomplete. Then systems that stopped guessing the next word and started guessing the next sentence, the next page, the whole program. Today’s models are vastly more capable. I don’t say this to diminish them; they are remarkable. But strip the layers back and the core is the same: predict what comes next.

And yet that prediction is not some hidden spark. It is the effect of an architecture, a stack of algorithms doing exactly what they were built to do. The model predicts because that is what the architecture produces, not because it wants anything at all. To be fair, the result can be uncanny. Give it enough about you and a typical-enough thought, and it will guess your next one, or even offer an idea you hadn’t reached yet. But it is still imitation, not anything genuinely its own. It mimics human thinking, and mimics it impressively, which is exactly why “intelligence” feels like the wrong word to me. We have built extraordinarily powerful machines that copy the intellectual qualities of people, and they run on an old human trick: fake it till you make it. A model can continue your thought. Can it ever begin one? It unfolds the question. We pose it.

Still, the more time I spend with these systems, the more I notice how much we resemble them. Don’t we learn in a similar way? We absorb language, books, films, songs, school, the endless talk around us, and then reproduce other people’s ideas as our own, often without realising it. Put that way, the human and the model start to look uncomfortably alike.

What a machine can’t do

Which brings me back to my student. If my only value as a teacher were to store and repeat information, a book could have done my job ages ago. What I do is harder to automate. A model can produce a flawless explanation. It cannot want you to understand.

My student wanted to know whether the machine would replace her and me. I think the better question is what it enables us to do. And that, in the end, is what it is: a tool. Like the technologies before it, it speeds work up, raises its quality, and opens up new kinds of work, sometimes whole fields that didn’t exist yesterday. For now, I see more that it adds than it takes away.

Originally published on LinkedIn.