"Artificial intelligence can now produce acclaimed essays and support medical diagnoses with impressive precision, yet biological brains still outperform machines in one essential area: flexibility. Humans can absorb new information and adapt to unfamiliar situations with very little effort. People can jump into new software, follow a recipe they have never tried before, or learn the rules of a game they have just discovered, while AI systems often struggle to adjust in real time and to learn effectively “on the fly.”
A new study from Princeton neuroscientists offers insight into why the brain excels at this kind of rapid adjustment. The researchers found that the brain repeatedly draws on the same cognitive “blocks” when performing different types of tasks. By recombining these blocks in new ways, the brain can quickly generate fresh behaviors.
“State-of-the-art AI models can reach human, or even super-human, performance on individual tasks. But they struggle to learn and
People often learn something new by building on related abilities they already have. Someone who knows how to maintain a bicycle, for example, may find motorcycle repair easier to pick up. Scientists refer to this process of assembling new skills from simpler, existing ones as compositionality.
“If you already know how to bake bread, you can use this ability to bake a cake without relearning how to bake from scratch,” said Sina Tafazoli, Ph.D., a postdoctoral researcher in the Buschman lab at Princeton and lead author of the new study. “You repurpose existing skills — using an oven, measuring ingredients, kneading dough — and combine them with new ones, like whipping batter and making frosting, to create something entirely different.”
Compositionality is considered central to human flexibility.
These cognitive Legos could explain why humans can learn new tasks so rapidly. Instead of generating each behavior from the ground up, the brain reuses existing components and avoids redundant work, something current artificial intelligence systems generally lack." SciTechDaily
