Hajra Naz
Despite rapid advances in artificial intelligence, the human brain continues to outperform computers in one crucial area: the ability to transfer knowledge from one task to another. A new study sheds light on how the brain manages this kind of flexible learning — and why even today’s most advanced AI systems still struggle to replicate it.
Studying Cognitive Flexibility Through Monkeys
The research was led by a team at Princeton University, but instead of testing humans, the scientists studied rhesus macaques, whose brain structure and cognitive abilities closely resemble our own.
The monkeys were asked to complete a series of related tasks involving identifying shapes and colors displayed on a screen. They signaled their answers by looking in specific directions. As the animals worked through the tasks, researchers monitored their brain activity to see how different regions were activated and reused.
‘Cognitive Legos’ Power the Brain’s Adaptability
Brain scans revealed that the monkeys weren’t creating entirely new neural pathways for each task. Instead, they relied on reusable clusters of neurons — what the researchers described as “cognitive Legos.” These building blocks could be rearranged and recombined to handle new challenges, allowing the brain to adapt quickly without starting from scratch.
This modular approach highlights a form of neural flexibility that current AI models lack. While machines can excel at narrowly defined problems, they often fail when asked to apply what they’ve learned to unfamiliar situations.
Why AI Still Struggles Across Multiple Tasks
“State-of-the-art AI models can reach human, or even super-human, performance on individual tasks,” said neuroscientist Tim Buschman of Princeton University. “But they struggle to learn and perform many different tasks.”
The study suggests that the brain’s strength lies in its ability to reuse and reorganize existing cognitive components rather than relearning everything from the ground up.
The Role of the Prefrontal Cortex
The researchers found that many of these cognitive building blocks were concentrated in the prefrontal cortex — a brain region associated with higher-level functions such as planning, decision-making, and problem-solving. This area appears to play a central role in enabling cognitive flexibility.
Interestingly, when certain cognitive blocks weren’t needed, their activity decreased. This suggests the brain can temporarily “store away” unused neural components, allowing it to focus more efficiently on the task at hand.
A Brain That Functions Like Software
Buschman likened these cognitive blocks to functions in a computer program. One group of neurons might process color, for example, while another translates that information into an action. By chaining these components together in different ways, the brain can perform complex tasks step by step.
This structure helps explain how monkeys — and likely humans — can tackle unfamiliar problems by drawing on existing knowledge, something artificial intelligence still finds difficult.
Implications for AI and Brain Health
The findings could eventually influence how AI systems are trained, offering clues on how to reduce a common problem known as catastrophic forgetting — where neural networks lose old skills as they learn new ones. The research may also prove valuable in developing treatments for neurological and psychiatric conditions that affect a person’s ability to apply learned skills in new contexts.
Why Flexibility Still Gives Humans the Edge
While constant task-switching isn’t ideal for the brain, the ability to transfer skills across situations provides a powerful shortcut for learning and adaptation.
“If the brain can reuse representations and computations across tasks,” the researchers concluded, “it can rapidly adapt to changes in the environment — either by learning from feedback or by recalling relevant knowledge from long-term memory.”
At a fundamental level, these “cognitive Legos” show why human intelligence remains more flexible and resilient than today’s AI — and why our brains still have one clever trick machines haven’t mastered yet.
