Most current technologies rely on mimicking the electrical activity of neurons, overlooking the overall structure of neural elements and their role in information processing. This innovative brain-like artificial system, detailed in a paper published in a prestigious scientific journal, emulates the organization of synapses (connections between neurons) and the branched structure of dendrites (projections extending from the neuron body).
The new device, known as the “dendristor,” replicates the computations performed by dendrites by employing the physical properties of multi-gate transistors coated with a doped sol-gel film. This film mimics dendritic branches, enabling ions to move similarly to their counterparts in neural dendrites, thereby modulating the transistor current to reflect changes in dendritic membrane potential.
While previous research in neuromorphic computing focused on synapse processes related to learning and artificial generation of neural impulses, this new approach faithfully reproduces dendritic neuron connectivity, showcasing remarkable energy efficiencies. Unlike conventional artificial neural networks, this system demonstrates the potential to detect movement using fewer neurons.
To read more, click here.