Artificial intelligence can write award-winning essays and diagnose diseases with impressive accuracy, yet in one crucial domain, biological brains still outperform AI: flexibility. Humans effortlessly adapt to new information and unfamiliar challenges—whether learning new software, following a recipe, or picking up a game—while AI systems struggle to adjust on the fly.
A new study by Princeton neuroscientists sheds light on why the brain has this edge. The research reveals that the brain reuses the same cognitive “building blocks” across multiple tasks, allowing humans to rapidly assemble new behaviors from familiar components.
“State-of-the-art AI can match or surpass human performance on individual tasks, but it struggles to perform many tasks simultaneously,” said Tim Buschman, Ph.D., senior author and associate director of the Princeton Neuroscience Institute. “The brain’s flexibility comes from reusing cognitive components across tasks. By snapping together these ‘cognitive Legos,’ it can generate entirely new behaviors.” The study was published on November 26 in Nature.
Reusing Skills to Tackle New Challenges
Humans often learn new skills by repurposing what they already know—a principle called compositionality. For example, knowing how to bake bread can make learning to bake a cake faster, since many skills—using the oven, measuring ingredients, kneading dough—transfer directly. “You can combine existing abilities with new ones, like making frosting, to create something entirely different,” explained Sina Tafazoli, Ph.D., lead author and postdoctoral researcher in the Buschman lab.
READ ALSO: https://www.modernmechanics24.com/post/soft-composite-symmetry-breaks-smart-tech
To study how the brain achieves this, Tafazoli trained two male rhesus macaques to perform three related categorization tasks while monitoring their brain activity. The tasks required judging whether a colorful, ambiguous blob resembled a bunny or the letter “T,” or whether it was more red or green. The monkeys indicated their choices by looking in different directions, with each task sharing certain elements with the others. This setup allowed researchers to test whether the brain reused neural patterns—the cognitive building blocks—across tasks with shared components.
Cognitive Legos Build Flexibility
Analysis showed that the prefrontal cortex, a region involved in higher cognition, contained reusable neural activity patterns across tasks. Buschman described these as the brain’s “cognitive Legos”—modular components that can be flexibly combined to perform new tasks. “Think of a cognitive block like a function in a computer program,” he said. One block might calculate color, another drive an eye movement; together, they enable task execution. Switching tasks simply involves recombining the relevant blocks.
WATCH ALSO: https://www.modernmechanics24.com/post/robot-dogs-explosives-china-battle-drill
The study also found that the prefrontal cortex suppresses blocks that aren’t currently needed, helping the brain focus on the task at hand. “You have to compress some abilities to prioritize what’s important,” Tafazoli explained. Focusing on shape categorization, for instance, temporarily diminishes color encoding because the goal is shape, not color.
Implications for AI and Medicine
These cognitive Legos may explain humans’ rapid learning and ability to generalize knowledge. AI systems, by contrast, often suffer from catastrophic interference, where learning new skills overwrites previous knowledge. “If an AI learns to bake cookies after mastering cake, it may forget how to bake a cake,” Tafazoli said. Incorporating compositionality into AI could enable machines to continually learn without forgetting prior skills.
READ ALSO: https://www.modernmechanics24.com/post/tum-startup-reinventing-rocket-pressure-tanks
The findings may also inform therapies for neurological and psychiatric disorders, such as schizophrenia or OCD, which often impair flexible use of known skills. “Understanding how the brain recombines knowledge could help restore the ability to shift strategies, learn new routines, and adapt to change,” Tafazoli said. Ultimately, these insights reveal not only why humans remain superior in flexible thinking but also how the brain’s modular approach could inspire AI and clinical interventions.
