Despite rapid advances in artificial intelligence in recent years, the humble human brain still outperforms computers in its ability to transfer skills and learn entire tasks. New research reveals how we might do this.
The researchers in this new study, led by a team at Princeton University, did not actually conduct experiments on humans, but instead used rhesus monkeys, an animal very similar to us in terms of biology and brain function. (macaw).
These monkeys were asked to identify shapes and colors on a screen and respond by looking in a specific direction. While this was happening, brain scans were used to check for overlapping patterns and common areas of activity in the animals' brains.
Related: Computers made from human brain tissue are coming. Are you ready?
These scans showed that the monkeys' brains used different blocks of neurons – “cognitive Legos” in the words of the researchers – throughout the task. Existing blocks can be reused and recombined across new tasks, exhibiting neural flexibility that even the best AI models can't compete with.
“Cutting-edge AI models can reach human or even superhuman performance on individual tasks,” said Tim Bushman, a neuroscientist at Princeton University. “But they have a hard time learning and performing many different tasks.”
“We found that the brain is flexible because it can reuse cognitive building blocks for different tasks. By combining these 'cognitive Legos', the brain can construct new tasks.”
As you can see in the video below, the animals had to identify shapes and colors in three separate but related tasks, and the animals had to continuously learn and apply what they knew from one task to the next.
frame border=”0″ permission=”accelerometer; autoplay; clipboard write; encrypted media; gyroscope; picture-in-picture. web-share” referrerpolicy=”strict-origin-when-cross-origin” allowedfullscreen>The cognitive Lego blocks the researchers identified were concentrated in the prefrontal cortex of the brain. This region is associated with higher-order cognition, such as problem solving, planning, and decision making, and appears to play an important role in cognitive flexibility.
The researchers also found that when a particular cognitive block was not needed, the activity of that block decreased, suggesting that the brain can file away neural legos that are not immediately needed to focus on the task at hand.
“I think about cognitive blocks, like functions in a computer program,” Bushman says.
“One set of neurons may discriminate colors, and their output can be mapped to another function that drives behavior. This organization allows the brain to perform a task by performing each component of the task in sequence.”
It explains how monkeys, and perhaps humans, can adapt to challenges and tasks never seen before and use existing knowledge to tackle them. This is something that artificial intelligence in its current form struggles with.
Further down the line, the researchers suggest their findings could help train AI to be more adaptable to new tasks. Their research could also help develop treatments for neurological and psychiatric disorders, where people struggle to apply skills to new environments.
For now, these cognitive Legos demonstrate, at a fundamental level, how the human brain is more flexible and adaptable than AI models that exhibit so-called catastrophic forgetting. This weakness means that neural networks cannot learn sequential tasks without forgetting how to perform the last task they were trained on.
Task switching isn't necessarily good for your brain, but applying what you know from one task to another can be a useful shortcut.
“If the brain is able to reuse representations and computations across tasks, as our results suggest, then it may be able to rapidly adapt to changes in the environment by learning appropriate task representations through reward feedback or by recalling them from long-term memory,” the researchers conclude.
This study nature.
