Over the course of a day, the brain accumulates impressions the way a room accumulates stuff. At night, during deep sleep, it doesn't just rest: it tidies up: it sorts through the day's events and decides what to keep and what to toss. Scientists built a computer model of the brain that learns from a few examples during the day and 'replays' them at night, as if dreaming. The model had connections within each layer of nerve cells and between layers, like shelves and drawers in a room.
It turned out that reshaping the connections between layers during sleep is key to solid memory. When those connections stayed fixed, the model did worse at recognizing images after sleep. But when the between-layer connections were updated, memory became sharper and more precise: the brain seemed to put things on the right shelves rather than piling them in a heap.
This nightly cleanup isn't just about saving energy: it's a way to make memory more reliable. The brain doesn't erase everything indiscriminately; it picks out what matters, strengthening useful connections and weakening random ones. That's why we remember important things in the morning, while trivial details fade away.
This finding hints at how to build brain-like computer systems that aren't energy hogs. If you let them 'sleep' and reorganize their connections, they could learn more accurately and consume less electricity, just like a brain that tidies up overnight and works more smoothly in the morning.
During sleep, the brain cuts energy use by tens of percent, like unplugging every appliance in the house except the fridge and the bedside lamp.