Where did the liberated brain go?
Last week, a friend said something over dinner that sounded obvious, but I kept thinking about it for days.
“Have you noticed that when we run into a problem now, our first thought is no longer ‘Let me think about it,’ but ‘Let me ask’?”
I said yes.
Last week, OpenAI released GPT-5.5 Instant with a striking figure attached: its hallucination rate was down 52.5%. People in AI kept repeating the number, with the implication that AI was finally becoming trustworthy.
Looking at that number, I thought of another question for which nobody has data:
After AI gives me an answer, how much less often do I seriously question it?
No one tracks that. No one cares.
In the past, when I ran into a problem, I would let it turn over in my head for a while. Often, that process went nowhere. I was just shuffling what I already knew, like someone wandering around a room without a destination.
But occasionally, in the middle of that apparently pointless rumination, an unexpected idea would pop up. A connection, a surprise, a little feeling of “wait, what if…?”
I have not felt that in a long time.
I have an unfinished theory: the brain is a little like a laid-off worker. The factory is still open, the machines are still running, but the orders are gone. You think it is resting, when really it just has nothing to do.
This is not a new problem. After calculators became common, our mental arithmetic got worse. After GPS became standard, many people who grew up in cities lost their sense of direction. Usually, an essay gets to this point and says, “But we traded those abilities for something more important.”
I would like to say that too. But I cannot think of what, exactly, we got in exchange.
Where did those “saved” cognitive resources go? Some say we can use them for higher-level thinking. That sounds right. But I took a look at my own life: most of the time I saved went to short videos.
What bothers me most is a tiny detail.
When I come across a new word now, I no longer try to guess what it means in my head first. I used to guess, get it wrong, look it up, and remember the word because of that moment of being wrong. The faint embarrassment of “Oh, so that is not what it means” was the best memory hook.
Now I ask AI directly. It answers. I nod. Then I forget.
AI’s hallucination rate is falling. So is the amount I remember.
I am not saying AI is bad. That would be silly. I am saying convenience has a cost, and the tricky thing about that cost is that it usually takes years to become visible.
No one predicted that “people will stop walking” when cars first became common. But we did stop. We even came to think it was natural, and to feel proud of it.
One day, I forced myself to answer a question without using AI. It took about 20 minutes. The conclusion was not great. But I noticed that I had drifted off three times along the way, thinking about three completely unrelated things. One of them reminded me of something I had long forgotten, and it lifted my mood for a long time.
Mind-wandering is a human privilege. AI does not daydream. There is no grand lesson in that, just a small reminder: do not hand every question over. Some detours are worth taking yourself.
The hallucination rate dropped by 52.5%.
When was the last time an answer I came up with myself surprised me?
I asked AI that question. It gave me a detailed answer, cited three studies and concluded that “it depends on the person’s cognitive habits and the way they use AI.”
I finished reading it, then forgot.