Study Finds 1 Human Neuron Matches Deep Neural Networks as a Biological Computer
Updated
Updated · ScienceDaily · Aug 13
Study Finds 1 Human Neuron Matches Deep Neural Networks as a Biological Computer
3 articles · Updated · ScienceDaily · Aug 13
Summary
A PNAS study found a single human cortical neuron can perform computations far beyond a simple on-off signal, behaving more like a tiny biological computer.
Hebrew University researchers reached that result by pairing computer modeling with AI, measuring how complex an artificial neural network had to be to mimic one neuron's input-output behavior.
Human cortical cells outperformed comparable neurons in other mammals because their branching dendrites and electrical properties support sophisticated processing of incoming information.
The findings challenge the idea that human intelligence depends mainly on the brain's nearly 100 billion neurons and their connections, pointing instead to the power of individual cells.
That framework could also guide brain-inspired AI, suggesting future systems may use deeper, more capable artificial units modeled on biological neurons.