WHILE artificial intelligence (AI) can master specific tasks, from writing essays to aiding medical diagnosis, it still lacks the mental flexibility of the human brain. We adapt to new software, recipes, or games with relative ease, a capacity that eludes most AI systems designed for singular purposes.

A new study from Princeton University, published in Nature, reveals a neural mechanism behind this advantage: the brain reuses shared cognitive “blocks” across different situations, assembling them like Legos to form new behaviors.

“State-of-the-art AI can reach super-human performance on individual tasks but struggles to learn many different ones,” said senior author Tim Buschman, Ph.D. “We found the brain is flexible because it reuses components of cognition. By snapping together these ‘cognitive Legos,’ it builds new tasks.”

This ability — termed compositionality — lets us apply known skills to new challenges. “If you can bake bread, you reuse those skills (using an oven, measuring) and combine them with new ones (whipping batter) to bake a cake,” explained lead author Sina Tafazoli, Ph.D.

To explore its neural basis, researchers trained monkeys on three visual tasks, requiring them to categorize ambiguous shapes or colors. Though each task had unique rules, they shared core components.

By recording brain activity, the team discovered that the prefrontal cortex, a region critical for complex thought, uses recurring neural patterns as cognitive building blocks. These “Legos” were reused whenever tasks shared features. This compositional reuse was distinct to the prefrontal cortex and not seen elsewhere.

The brain also strategically quiets unneeded blocks, preventing overload and maintaining focus on the current goal. This efficient system of recombining and suppressing pre-existing components allows for rapid learning without starting from scratch — a key insight for developing more adaptable AI.

(SD-Agencies)