A new study by neuroscientists at Princeton University has revealed a key mechanism by which the brain maintains its advantage in the face of artificial intelligence (AI): the brain reuses the same cognitive "modules" in different tasks, and by combining and reorganizing these modules, much like "assembling building blocks," the brain can rapidly construct new behavioral patterns. This research was published in the latest issue of the journal *Nature*.
While the most advanced AI models can match or even surpass human performance on a single task, they perform poorly when learning and performing multiple different tasks. This is because the human brain still maintains an advantage in a key aspect: cognitive flexibility. For example, humans can adapt to new information or unfamiliar challenges relatively easily, such as learning new software, cooking according to a recipe, or quickly mastering a new game, while AI struggles to achieve such "instant" learning.
Research has found that the brain's flexibility stems from its ability to reuse cognitive components, assembling them like building blocks. For instance, someone skilled at repairing bicycles can readily apply their skills when learning to repair motorcycles. However, the exact mechanisms by which the brain achieves this cognitive flexibility remain unclear and lack consistent, comprehensive evidence within the scientific community.
The team trained two male rhesus monkeys to complete three related classification tasks and recorded their brain activity during the process. These tasks were similar to judging whether a blurry image on a screen looked more like a rabbit or the letter "T," or more like red or green. The blurriness of the images varied, sometimes obvious, sometimes subtle. The monkeys expressed their judgments by looking in different directions. The key to the experimental design was that while each task was different, they shared certain cognitive elements, allowing the team to examine how the brain repeatedly uses the same neural activity patterns.
Through multi-brain region analysis, the team discovered that the prefrontal cortex (the brain region responsible for higher cognitive functions) contains several reusable activity patterns, which correspond to different "cognitive building blocks." The brain constructs new behaviors by flexibly assembling these "building blocks."
Furthermore, the prefrontal cortex also suppresses the activity of certain cognitive modules when they are not needed, thus helping the brain focus on the current task. This is because cognitive resources are limited, and the brain ensures that the primary target task is not disturbed.
The team claims that this combinatorial learning approach may be key to humans' ability to efficiently learn new skills without forgetting old ones. In contrast, current AI systems often suffer from "catastrophic interference"—learning new tasks often overwrites previous memories. Introducing this modular, reconfigurable mechanism of the human brain into AI might enable the development of intelligent systems that can continuously learn without forgetting. Furthermore, this discovery has clinical implications for understanding certain mental illnesses and brain injuries.
[Editor's Note]
Many characteristics of the human brain are beyond the reach of current AI. If you have children around you, you'll know that even two- or three-year-old toddlers can assemble toys by following diagrams or learn to use a mobile phone without instruction. This research found that the prefrontal cortex of our brains flexibly combines existing standardized "cognitive modules" like building blocks to quickly respond to new challenges. If this mechanism can be integrated into AI design, it might be possible to develop intelligent systems that can continuously accumulate experience, taking an important step towards general artificial intelligence. The research also suggests that some mental illnesses may be related to impaired cognitive module reorganization mechanisms.
