Exploring Neuromorphic Nanowire Networks for Human-Like Intelligence

Recent advancements in generative AI models such as ChatGPT and DALL-E have made it feasible to produce vast amounts of human-like, high-quality creative content from a simple set of prompts. However, despite being highly capable, AI systems are not structured like human brains and do not learn in the same way. They are also not as adaptable and lack human-like memory capabilities.

The Study

A study published in Science Advances explores non-biological systems that are more like human brains. The research focuses on a system that uses a network of “nanowires” to mimic the neurons and synapses in the brain. These nanowires are tiny wires about one thousandth the width of a human hair, made of a highly conductive metal typically coated in an insulating material like plastic. Nanowires self-assemble to form a network structure similar to a biological neural network.

Learning and Memory

The study demonstrates that we can selectively strengthen and weaken synaptic pathways in nanowire networks, similar to “supervised learning” in the brain. The output of synapses is compared to a desired result, and then the synapses are strengthened or pruned accordingly. The researchers expanded on this result by showing that they could increase the amount of strengthening by “rewarding” or “punishing” the network, inspired by “reinforcement learning” in the brain.

The researchers also implemented a version of a test called the “n-back task,” used to measure working memory in humans. The network “remembered” previous signals for at least seven steps. Interestingly, seven is considered the average number of items humans can keep in working memory.

Synthetic Intelligence

Nanowire networks are different from the artificial neural networks used in AI, but they may lead to so-called “synthetic intelligence.” Although human intelligence is still likely far from being replicated, this research on neuromorphic nanowire networks shows that it is possible to implement features essential for intelligence, such as learning and memory, in non-biological, physical hardware. Perhaps one day, a neuromorphic nanowire network could learn to have conversations that are more human-like than ChatGPT and remember them.