Generative AI models (like ChatGPT and DALL-E) have enabled the production of human-like creative content from simple prompts, even so, these systems are not structured like human brains and lack similar learning and memory capabilities; but in a recent study published in Science Advances, researchers explored non-biological systems that mimic the brain’s structure and functionality, and they found that self-organizing networks of tiny silver wires — called nanowires — appear to learn and remember in a way similar to the “thinking hardware” in our brains.
The nanowires self-assemble to form a network structure similar to a biological neural network, with each wire coated in a thin insulating layer. Electrical signals applied to the nanowires cause ions to migrate across the insulating layer and into a neighboring wire, resulting in synapse-like electrical signaling.
The researchers found that synaptic pathways in the nanowire networks can be selectively strengthened (or weakened), similar to supervised learning in the brain. They also implemented reinforcement learning — where the network was rewarded or punished based on its output — leading to improvements in memory performance; the networks demonstrated memory for at least seven steps, similar to the average number of items humans can keep in working memory. The formation of synaptic pathways in the nanowires depended on past activation, a phenomenon known as metaplasticity. Via TheConversation.
While human intelligence is still a long way from being replicated, the study shows it is possible to implement essential intelligence features, like learning and memory, in non-biological physical hardware. This research on neuromorphic nanowire networks may lead to “synthetic intelligence” and the ability to have more human-like conversations and remember them. Sounds exciting!
Filed in AI (Artificial Intelligence).. Read more about