Our Brains Still Outperform AI with One Simple Trick
Even as artificial intelligence makes rapid leaps forward, the human brain remains adept at transferring skills and learning across different tasks. A recent study sheds light on how we might accomplish this so effectively.
Led by researchers at Princeton University, the team didn’t conduct experiments on humans. Instead, they studied our closest biological relatives: rhesus macaques (Macaca mulatta). In a series of tasks, the monkeys were asked to identify shapes and colors on a screen and to direct their gaze in specific ways to indicate their choices. While the tasks unfolded, researchers used brain imaging to look for overlapping activity patterns and shared neural regions.
These scans revealed that the monkeys’ brains recruit distinct neural blocks—cognitive building blocks, as the researchers describe them—across different tasks. The same blocks can be repurposed and recombined to handle new challenges, illustrating a level of neural flexibility that current AI models struggle to match.
“State-of-the-art AI models can achieve human or even superhuman performance on isolated tasks,” notes Tim Buschman, a Princeton neuroscientist involved in the study. “But they falter when it comes to learning and performing across many varied tasks.”
The team emphasizes that the brain’s strength lies in its ability to reuse components of cognition across tasks. By snapping these cognitive building blocks together, the brain can construct new solutions for new tasks.
In the video accompanying the study, you can see the monkeys tackling three related tasks that require them to continually apply prior knowledge to new situations. The cognitive blocks reside mainly in the prefrontal cortex, a region tied to higher-order thinking such as problem-solving, planning, and decision-making. This area appears to play a pivotal role in cognitive flexibility.
Moreover, when certain blocks aren’t needed, the brain dampens their activity, suggesting it can store away unused components to focus on the current task more efficiently.
Buschman compares a cognitive block to a computer function: one neuron group may discriminate color, and its output can feed into another function that drives action. This modular organization enables the brain to perform complex tasks by sequentially deploying each component.
This flexibility helps explain how monkeys—and likely humans—can adapt to new challenges and tasks they’ve never encountered before by leveraging existing knowledge. It’s a capability that today’s AI, with its tendency toward catastrophic forgetting, struggles to replicate.
Looking ahead, the researchers propose that understanding these cognitive Lego blocks could guide the development of AI systems that are better at adapting to unfamiliar tasks. The findings might also inform approaches to treating neurological and psychiatric conditions where people have difficulty applying learned skills to new contexts.
For now, these neural building blocks illustrate, at a fundamental level, why human cognition remains more versatile than current AI models. Unlike neural networks, which can forget earlier tasks as they learn new ones, the brain’s architecture supports broad, cross-task reuse of knowledge.
While multitasking isn’t ideal for brain health, applying prior knowledge across tasks offers a practical shortcut in real-world problem solving.
If these results hold, future AI could become more adaptable by reusing representations and computations across tasks, enabling quicker adaptation to changing environments through reward-based learning or memory recall. The study documenting these insights appears in Nature.
