Korean researchers have successfully built artificial neurofiber transistors with structure and functionalities comparable to the human brain and nervous system and can be utilized as a neural network.
Developments in AI-based technology have resulted in an exponential rise in the quantity of data accessible for computer processing. Current computer systems frequently process the data, requiring a considerable amount of time and resources to analyze vast amounts of data. To address such difficult challenges, a shift to a new computing model is necessary. Large-scale calculations must be conducted in a way comparable to processing data in the human brain, which necessitates research into devices that can serve as neurons and synapses. Scientists are now working to create energy-efficient neuromorphic computing systems and hardware designed to process vast quantities of data by replicating the structure and processes of the human brain.
Unlike previously produced devices that operate either as synapses or neurons, the artificial neurofiber transistor can replicate both neuronal and synaptic characteristics. By linking transistors in arrays, a structure simulating a neural network could be easily constructed. Biological neurons contain fiber branches that may accept numerous inputs at the same time, and signal transmission is facilitated by electrically induced ion migrations.
The artificial neurofibers were created by researchers utilizing fibrous transistors that had previously been invented in 2019. The team had created memory transistors that, like the synapses, and recall the intensities of the applied electrical impulses and transfer them with the help of redox reactions among the semiconductor channels and ions inside the insulators in response to electrical inputs from neurofiber transistors. Such artificial neurofibers also replicate neurons' signal summing capability. Using the artificial neurofibers, the researchers constructed an artificial neural network system with 100 synapses and tested the device's stability using voice recognition studies. After training with voice samples, the designed neural network achieved a recognition rate of 88.9 %.