Who gets to learn the slow way?

I was in the audience at a symposium when Erik Scherder, a neuropsychologist, said it plainly: the more we use AI, the dumber we’ll get. The room did the thing rooms do when someone respected says something uncomfortable and probably true. A small stir, then quiet.

What surprised me was where my own head went. Not to myself. I’m still learning something most days. But I learned how to think before the tools showed up to do it for me. My head went to the row behind me, to the people just starting out. I couldn’t shake the sense that the warning was landing on the wrong people first.

So I went looking for why. The first thing I found is that Scherder’s claim isn’t really about AI at all. Pressed on it, what he points to is the lack of thinking. Not the tool, the step the tool lets you skip. That’s a sharper claim, and an older one. A search engine still makes you choose, weigh, decide what fits. A model that writes the whole thing for you takes the step away. So “AI makes you dumber” is really “skipping the thinking makes you dumber,” and that was true long before any of this arrived.

There’s evidence for it now, which there wasn’t a few years ago. A study out of MIT last year wired people up while they wrote essays. Some used an AI assistant, some a search engine, some nothing at all. The ones leaning fully on the model showed the weakest brain activity, and most of them couldn’t quote back work they’d finished minutes earlier. But the part most people skip is what happened when the groups swapped. The people who had written unaided first, and only then brought the tool in, re-engaged: their recall held up and the activity came back, nearer the search-engine users than the ones who’d leaned on the model all along. Same tool, opposite result. What separated them was the order: thinking first, or tool first.

I should admit my own hand here. I do the thinking and hand the writing to the machine. Which, given what I just said, ought to worry me more than it does. The test I’ve settled on is the friction when I read a draft back. If editing it makes me think “no, the real point is sharper than that,” the thinking wasn’t done, and the writing is still doing work on me. The day the drafts go down easy and I just nod them through is the day I’ve crossed over.

For me the warning is easy to follow. Build the skill first, then bring in the tool. I can do that, because I already have the skill. Say the same sentence to someone who’s twenty-three and it stops being advice and turns into a paradox. They meet the tool before there’s any skill to build first, in the exact window the skill was supposed to form.

And the thing the tool takes away, for them, was more than the output. The clumsy first memo. The spreadsheet built by hand that breaks somewhere obvious. The analysis that comes back from a senior covered in red ink. Everyone treats that work as the price of having a junior around, slow and a little painful and worth it later. But the deliverable was never the point. The correction was. The grunt work was the apprenticeship wearing the costume of cheap labor, and if you automate the labor you quietly automate the apprenticeship with it.

What makes it hard to stop is that every step toward it is reasonable. Why pay a junior to slowly produce something the tool produces faster and cleaner? No single manager is wrong to ask. But the slow version was building the person who, ten years out, can look at the tool’s confident answer and see what’s wrong with it. That payoff never shows up on this quarter’s numbers. It shows up in the whole field’s supply of people who can still think, years later, where no one owns it and no one is billed for it. So each firm does the sensible thing, and the sensible things add up to a shortage nobody chose.

So the warning stands, and I’ll keep giving it: build the skill before you hand it off. I just keep walking into the edge of it. I got mine the slow way, before there was a tool to skip the slow way with. I don’t know how the person behind me gets theirs. The work I built mine on is exactly the work we’re most eager to automate now. Maybe there’s a new apprenticeship in here that I can’t see yet, because no one has had to build it before. Or maybe we’re spending the thing that made us worth hiring and calling it efficiency. I can’t tell yet which. That’s the part I can’t put down.