Skip to content
IQBAL.LOG
QUERY_ARCHIVE

AI // Applied AI & machine learning, explained twice.

AI Takes the Effort. The Effort Was the Point.

I went down a rabbit hole about what AI does to how we learn. Outsource the effort and you lose the faculty — unless you keep the thinking for yourself.

24 Jun 2026 4 min read AI-assisted

I’ve got two kids now, and a worry I can’t quite shake: what is AI doing to how they’ll learn? One weekend I followed it down a rabbit hole.

It started with something I half-remember reading years ago — a class of children who couldn’t follow simple directions without Google Maps, even with the street signs right in front of them. I couldn’t find the source, so treat that as a story, not a citation. But it sent me looking, and what I found is well documented.

TL;DR // Key Takeaways ANSWER_FIRST
  • 01 The harm isn’t AI doing the work — it’s AI doing the thinking for you.
  • 02 Lean on GPS and your spatial memory measurably fades: the brain’s mapmaker does less.
  • 03 People who wrote essays with ChatGPT showed the weakest brain engagement — and couldn’t quote what they’d “written”.
  • 04 You atrophy what you outsource. Hand over the drudgery; keep the judgement.

Your brain runs on use-it-or-lose-it

Start with navigation, because the evidence there is the clearest. A 2020 McGill study looked at 50 regular drivers and measured how much they had leaned on GPS over their lives. The people with the heaviest GPS habit were worse at finding their way when they had to navigate without it. A smaller follow-up retested people three years later, and the ones who had used GPS most in the meantime had declined the fastest.

Here’s the part that matters. The heavy GPS users hadn’t started out with a poor sense of direction — which points to the uncomfortable reading: the GPS use came first, and the decline followed. The hippocampus is the brain’s mapmaker, and like a muscle it responds to load. Stop asking it to build maps and it gets worse at building them.

The same thing, but for thinking

Now swap navigation for thought. Last year, MIT’s Media Lab ran a study with the blunt title Your Brain on ChatGPT. Fifty-four people wrote essays in one of three ways: with ChatGPT, with a plain search engine, or with nothing but their own head. The researchers watched their brain activity with an EEG cap the whole time.

The result was a clean gradient. The people working unaided showed the strongest, widest brain engagement. The search-engine group sat in the middle. The ChatGPT group showed the weakest engagement of all — and when asked afterwards, 83% of them couldn’t quote a single line from the essay they had just produced. They’d offloaded the effort, and the effort was the point. There’s nothing to remember from work your brain never really did.

The reframe: you atrophy what you outsource

That is the through-line, and it is the whole point of the piece. The faculty you stop exercising fades, whether it’s finding your way home or thinking a problem through. The tool isn’t the villain. The villain is handing over the part that was doing you good.

We’re already doing it to ourselves

This is not only a worry about children. Watch how adults already work.

I keep noticing a loop. Someone writes something and sends it over. The person on the other end feeds it to AI for a summary, writes a reply, and that reply gets summarised too. At some point you realise nobody actually read anything. Two AIs are passing notes, and we’re the couriers.

It’s everywhere now — people summarising articles that were already short, getting the gist of a four-minute read just in case the four minutes were a waste. Some of that is fine. But what does not survive a summary is usually the part that made the thing worth reading: the one line that sticks, the point the writer took 600 words to earn, the understanding you only get from doing the reading yourself. We don’t read any more. We extract.

It isn’t inevitable

Here’s the hopeful part, and I want to be careful not to oversell it. The same MIT work had a smaller follow-up session that hinted at a way out: the people who wrote unaided first, and only then brought the AI in, kept more of their engagement than the people who leaned on it from the start. The order matters. Do the thinking, then use the tool to sharpen and speed it — not the other way round.

So I won’t keep AI away from my kids. Used well, it genuinely helps. But I want them, and honestly all of us, using it in a way where the tool takes the drudgery and leaves the thinking where it belongs. That has to be designed for, deliberately, and right now almost nobody is doing that.

It matters even more for children who learn differently — the ones the standard path already doesn’t fit. Used the right way, the right tool removes a real barrier. Used the wrong way, it quietly removes the learning along with it.

Let AI take the effort, never the thinking — because the effort was always where the learning lived.

Related Logs

Subscribe for updates

Occasional deep-dives on applied AI and systems at scale — delivered to your inbox.