New Crystal Technology Could Revolutionize Neuromorphic Computing and AI

Researchers at ITMO University have uncovered a significant advancement in neuromorphic computing by demonstrating that hybrid crystals can simulate the behavior of brain neurons when exposed to laser light. The study utilized crystals composed of copper and trimesic acid—materials noted for being environmentally safe and cost-effective to produce. When subjected to laser pulses at specific intervals, these crystals exhibited reactions similar to how neurons respond to neurotransmitters.

In experiments, a neural network built with these crystals was able to learn and recognize handwritten text more quickly and accurately than traditional neural networks. Remarkably, the crystal-based network achieved 100% accuracy with fewer training cycles. A key advantage of this technology is that all operations are conducted within a single crystal, unlike classical electronic devices where operations occur across multiple components. This integrated approach significantly reduces both energy consumption and processing time.

This discovery could address the limitations of current artificial intelligence algorithms that rely on conventional electronic platforms. While these platforms possess power comparable to the human brain, their operational principles hinder the full potential of neural networks. The application of crystals in neuromorphic computing presents new possibilities in artificial intelligence, potentially leading to the development of faster and more efficient machine learning systems.