The Brain’s ‘Cognitive Legos’: How Human Flexibility Outpaces AI
A new study published November 26 in the journal Nature reveals a key reason why humans excel at adapting to new situations while artificial intelligence often struggles: the brain’s remarkable ability to reuse core cognitive components, effectively building new skills from familiar “building blocks.”
Despite recent advances allowing artificial intelligence to perform tasks like writing essays and diagnosing diseases with increasing accuracy, a fundamental gap remains between human and machine intelligence. Humans demonstrate an innate flexibility, quickly mastering new software, recipes, or games – a skill that currently eludes even the most sophisticated AI systems.
Researchers at Princeton University have pinpointed a crucial element driving this difference: compositionality, the brain’s capacity to combine and recombine existing cognitive skills to tackle novel challenges. “State-of-the-art AI models can reach human, or even super-human, performance on individual tasks. But they struggle to learn and perform many different tasks,” explained a senior author of the study. “We found that the brain is flexible because it can reuse components of cognition in many different tasks. By snapping together these ‘cognitive Legos,’ the brain is able to build new tasks.”
Understanding Compositionality: Building Skills Upon Skills
This ability to leverage prior knowledge isn’t simply about efficiency; it’s fundamental to how we learn. Consider learning to repair a motorcycle after already knowing how to tune a bicycle. The existing knowledge base simplifies the process, allowing for faster comprehension and skill acquisition. As one researcher explained, “If you already know how to bake bread, you can use this ability to bake a cake without relearning how to bake from scratch.” The process involves repurposing familiar skills – using an oven, measuring ingredients – and integrating them with new ones, like whipping batter and making frosting.
Until now, the neurological basis for this flexible thinking remained somewhat elusive. To gain a clearer understanding, researchers trained two male rhesus macaques to perform a series of related visual categorization tasks while monitoring brain activity.
Visual Categorization and the Brain’s Adaptability
The monkeys were presented with colorful, balloon-like shapes on a screen and tasked with categorizing them based on shape (bunny vs. letter “T”) or color (red vs. green). The difficulty lay in the ambiguity of the images, requiring careful judgment to differentiate between categories. Monkeys indicated their choices by looking in specific directions on the screen.
The experimental design was carefully constructed to share key components across tasks. For example, some tasks utilized the same directional cues, while others required similar color categorization methods. This allowed researchers to observe whether the brain reused the same neural patterns when tasks shared common features.
The Prefrontal Cortex: A Hub for Cognitive Building Blocks
Analysis of brain activity revealed that the prefrontal cortex, a region associated with higher-level thinking and decision-making, contained recurring patterns of activity. These patterns emerged when groups of neurons collaborated toward a common goal, such as distinguishing colors. Researchers likened these patterns to the brain’s “cognitive Legos” – reusable building blocks that can be flexibly combined to produce diverse behaviors.
“I think about a cognitive block like a function in a computer program,” one researcher stated. “One set of neurons might discriminate color, and its output can be mapped onto another function that drives an action. That organization allows the brain to perform a task by sequentially performing each component of that task.”
For instance, when categorizing color, the brain assembled a block for color determination alongside a block guiding eye movements. When switching to shape categorization while maintaining the same eye movements, the brain simply activated the shape processing block with the existing eye movement block. This sharing of blocks was most prominent in the prefrontal cortex, suggesting its central role in compositionality.
The research also revealed that the prefrontal cortex actively suppresses irrelevant cognitive blocks when they are not needed. This selective activation and suppression allows the brain to concentrate on the most pertinent task, preventing cognitive overload. “The brain has a limited capacity for cognitive control,” explained a researcher. “You have to compress some of your abilities so that you can focus on those that are currently important. Focusing on shape categorization, for example, momentarily diminishes the ability to encode color because the goal is shape discrimination, not color.”
Implications for AI and Mental Health
These findings offer valuable insights into the differences between human and artificial intelligence. While AI often suffers from catastrophic interference – forgetting previously learned information when acquiring new skills – the brain’s compositional approach allows for continuous learning without erasing past knowledge. “A major issue with machine learning is catastrophic interference,” one researcher noted. “When a machine or a neural network learns something new, they forget and overwrite previous memories.”
Incorporating principles of compositionality into AI development could lead to more adaptable and human-like artificial systems. Furthermore, understanding these cognitive mechanisms could have significant implications for medicine. Many neurological and psychiatric conditions, including schizophrenia, obsessive-compulsive disorder, and brain injury, can impair the ability to apply existing skills in new situations. These challenges may stem from a disruption in the brain’s capacity to recombine cognitive building blocks.
“Imagine being able to help people regain the ability to shift strategies, learn new routines, or adapt to change,” one researcher said. “In the long run, understanding how the brain reuses and recombines knowledge could help us design therapies that restore that process.”
