Current artificial intelligence, despite its admirable ability to process enormous amounts of data, shows one enormous energy voracity, which grows with real-time processing. To overcome this obstacle, a group of scientists forged a physical neural network out of nanoscale silver wires.
This network, in addition to learning in real time, promises significantly higher efficiency. With an “on-the-fly” learning mode that approaches human learning, AI can evolve in ways we could only imagine until now.
The silver brain: how to build a metallic “mind”.
The new physical neural network, painstakingly woven from nanometer silver threads, was created to learn “on the fly.” These threads, despite being as thin as one thousandth the width of a human hair, form a structure that emulates the complexity and functionality of the neuronal networks of our brain, and represents a breakthrough in the field of neuromorphic computing.
The beating heart of this network lies in its response to electrical signals. Each intersection between the silver nanowires represents a synapse, shaping electrical flow in response to external signals, just as a biological synapse would. With tens of thousands of these synaptic intersections, the network can process and transmit information with unprecedented efficiency.
Real-time learning
Unlike conventional batch learning methodologies, this physical neural network takes in data in a continuous stream, learning and adapting in real time. This online learning approach reduces the need for bulky memories and high power consumption, typical of traditional AI.
The tests conducted demonstrated the ability of the silver neural network to recognize handwritten numbers and memorize sequences of digits, exhibiting real-time learning. But we are only at the beginning. The immense potential of these physical neural networks may just be scratching the surface of an ocean of unexplored possibilities. This is not just a technological advancement, but a promise of a future in which AI could evolve more analogously to human intelligence, opening doors to unimaginable horizons.
In short (Italian only)
The path to more efficient, greener AI may not be as far away as we thought. The promise of machines that learn and think more humanely is not a distant dream, but a tangible reality that may be within our reach far sooner than we dare imagine.
If you want to delve deeper into this important research, you can find everything in the paper published in Nature Communications.
