The world‘s fascination with artificial intelligence is relentless, driven in large part by its ability to handle massive amounts of data. But current AI models, which rely on energy-intensive artificial neural networks, falter with real-time data. In a revolutionary step, researchers have ventured away from conventional AI architectures, instead sculpting a physical neural network out of silver nanowires. This is reported in a new article in The Conversation written by Zdenka Kuncic, professor of Physics, and Ruomin Zhu, doctoral student, both from the University of Sydney.[1]
Emulate the human brain
Using nanotechnology, Kuncic and Zhu built networks out of silver nanowires, each as thin as one thousandth of a human hair. These nanowires spontaneously form random networks, eerily echoing the intricate networks of neurons in our brains. This innovative approach has its roots in neuromorphic computing, a field that seeks to mimic the brain functions of neurons in tangible hardware. When introduced to external electrical signals, nanowire networks adapt similarly to biological synapses.
Revolutionizing machine learning
The study, a collaboration between the University of Sydney and the University of California, Los Angeles, demonstrated the nanowire network’s ability for online machine learning. Unlike traditional machine learning that processes data in chunks, the online approach continuously transmits data, allowing the system to learn and adapt instantly. Notably, this method requires significantly less memory and energy.
Furthermore, the nanowire network demonstrated its proficiency in recognizing handwritten numbers and remembering digit patterns, underscoring its potential for brain-like learning.
I study
Nanowire networks (NWNs), an emerging category of neuromorphic systems, utilize unique material properties of the nanostructure. This study delves into how NWNs can be employed for online learning using image classification and sequential memory recall tasks. Achieving an impressive 93.4% accuracy, the results of this new study reveal the capabilities of spatiotemporal learning with NWNs, underscoring how memory can amplify learning. The study was published on Nature Communications.[2]
Insights
We built a ‘brain’ from tiny silver wires. It learns in real time, more efficiently than computer-based AIOnline dynamical learning and sequence memory with neuromorphic nanowire networks | Nature Communications (DOI: 10.1038/s41467-023-42470-5)
Share this article