CUHK-CityU Study Reveals Brain Multitasking Mechanism, Offers AI Training Insights
AM730 · 2 SOURCESabout 2 hours ago2 MIN
Summary
A joint research team from the Chinese University of Hong Kong (CUHK) Faculty of Medicine and City University of Hong Kong (CityU) has uncovered the neural mechanisms behind the brain's ability to learn multitasking. Published in the prestigious journal Neuron, the study reveals how the brain dynamically reorganizes neural resources to handle multiple tasks simultaneously, offering valuable insights for artificial intelligence development.
Key Points
- Researchers led by Prof. Ke Ya from CUHK and Prof. Yung Wing-ho from CityU conducted experiments on mice to understand multitasking mechanisms
- The team designed a dual-task paradigm where mice maintained continuous lever-pressing while making decisions based on auditory cues
- Using two-photon calcium imaging, they tracked individual neuron activity in the secondary motor cortex (M2) over weeks of training
- Key finding: the brain recruits task-specific neurons while gradually differentiating neural characteristics to reduce interference between tasks
- When M2 activity was suppressed, mice showed no improvement in multitasking despite training; reactivating M2 rapidly restored ability
- Deep learning tests using recurrent neural networks confirmed that preserving early coordination mechanisms accelerates multi-task learning
Why It Matters
The research demonstrates that efficient multitasking requires balancing neural coordination with representation separation—not complete resource sharing or strict compartmentalization . These findings could revolutionize AI training methodologies, enabling systems that learn multiple tasks more quickly and effectively, while also shedding light on cognitive deficits associated with neurological disorders .
The research demonstrates that efficient multitasking requires balancing neural coordination with representation separation—not complete resource sharing or strict compartmentalization . These findings could revolutionize AI training methodologies, enabling systems that learn multiple tasks more quickly and effectively, while also shedding light on cognitive deficits associated with neurological disorders .