Recent advancements in artificial intelligence (AI) have made it possible to produce human-like content with less effort. Though AI is capable of surpassing the skills of humans in tasks like big-data pattern recognition, they lack the ability to learn and memorize in the same way our brains do and are energy and resource-intensive. In response to this, researchers from the field of neuromorphics have recently discovered a new approach to creating a system that mimics the structure and functionality of biological neurons and synapses in non-biological systems.
In a study featured in Science Advances, scientists have used self-organizing nanowire networks to demonstrate the potential of artificial intelligence to learn and remember. The nanowires comprising the networks are made of highly conductive metals such as silver and are about one thousandth the width of a human hair. These wires are coated with an insulating material, like plastic, to serve as the neuronal layer. When stimulated with electrical signals, ions migrate across the nanowires’ insulating layer, creating synapse-like electrical pathways which are akin to the neurotransmitters of our brains.
The researchers were able to observe learning and memory, which are integral features of high-order cognitive functions. By strengthening and weakening the synaptic pathways of the nanowire networks, they were able to emulate a process known as “supervised learning.” They then improved their results by introducing “reinforcement learning” and were able to increase the amount of strengthen networking. To prove the efficiency of their system, the researchers implemented a test used to measure human working memory, the “n-back task.” In the n-back task, the nanowire networks were able to store memories for up to seven steps, the approximate amount of items humans can recall.
This research has laid the foundation for the potential of synthetic intelligence to replicate learning and memory. To develop these developments, neuroscientists must be able to construct their systems in a way that makes them more adaptable to dynamic and unpredictable environments, similar to the way our brains function. While much progress still needs to be made, this research emphasizes the potential for non-biological systems to come closer to replicating human intelligence in the near future.
