Scientists have developed a new type of electronic component that could help artificial intelligence systems retain old knowledge while learning new skills—a fundamental challenge that has limited the development of more capable AI. The innovation, detailed in a study published this month in Nature Communications, represents a significant advancement in neuromorphic computing, which aims to mimic the brain‘s efficiency and learning capabilities.
Unlike conventional computer chips that process information sequentially, these specialized components—called memristors—can process and store information simultaneously, similar to how neurons and synapses function in the human brain. What makes the new devices particularly promising is their ability to overcome a persistent problem in AI known as “catastrophic forgetting.”
To read more, click here.