Using neuroscience to develop artificial intelligence

Deep network is a brain-inspired model which is consist of multiple layers of neuron-like elements, connected by synapses. These models proved superior to previously known methods in central areas of AI research.

Comparisons of deep network models with empirical physiological, functional magnetic resonance imaging, and behavioral data have shown both similarities and dissimilarities between brain and new deep network models.

In comparisons with the primate visual system, similarities between physiological and model responses were identified. Another similarity is using reinforcement learning (RL), which is similar to reward signals in the brain that are used to modify behavior.

The superiority of human cognitive learning and understanding compared with existing deep network models may largely result from much richer and complex innate structures incorporated in human cognitive system. Training of current artificial intelligence models relies on large training datasets. Whereas biological systems often perform complex behavioral tasks with limited training based on pre-existing network structures.

To overcome these challenges, combination of experimental and computational approaches is necessary so that both neuroscience researchers and deep network developers can use it.

Reference