Published in


‘Brains Are Amazing’ — Neuroscientists Discover L2/3 Human Neurons Can Compute the XOR Operation

When talk of artificial neural networks began some fifty years ago the idea was to mimic the behaviour and function of the neurons in human brains — a premise that has more or less survived to this day. But new research now suggests scientists may have severely underestimated the power and potential of our neurons.

“Brains are amazing. Our lab demonstrates that single human layer 2/3 neurons can compute the XOR operation. Never seen before in any neuron in any other species…” The tweet from Humboldt University of Berlin Research Fellow and Neuroscientist Jaan Aru introduced the paper Dendritic Action Potentials and Computation in Human Layer 2/3 Cortical Neurons to a surprised neuroscience research community. Recently published in Science, the paper authors are from Humboldt, Charité Universitätsmedizin Berlin, and the Foundation for Research and Technology — Hellas (FORTH).

Why human layer 2/3 neurons?

The human brain has evolved an extraordinarily thick cortex with layers 2 and 3 (L2/3 of 6) particularly important in cognitive function. These L2/3 human cortical neurons form dendritic trees, which the researchers explain “largely determine the repertoire of transformations of the synaptic inputs to axonal action potentials (APs) at the output. Thus, they constitute a key element of the neuron’s computational power.”

It’s believed one of the reasons these human brain neurons’ abilities remained undiscovered is because previous knowledge of active dendrites was developed almost entirely from studies on rodents.

In their investigation of the dendrites of L2/3 neurons the researchers discovered waveform and effects on neuronal output fromdCaAPs, a class of calcium-mediated dendritic action potentials: “In contrast to typical all-or-none action potentials, dCaAPs were graded; their amplitudes were maximal for threshold-level stimuli but dampened for stronger stimuli. These dCaAPs enabled the dendrites of individual human neocortical pyramidal neurons to classify linearly nonseparable inputs — a computation conventionally thought to require multilayered networks.”

It had been assumed that tasks such as XOR could not be performed by a single neuron. In fact it takes a two-layer artificial neural network to compute XOR (Exclusive Or) — a basic logical operation that gives a true (1 or HIGH) output when the number of true inputs is odd.

The discoveries provide an exciting new perspective on how neurons work and how our brains process information. Says Oberlin College Neuroscience graduate Yujia Liu: “If the findings can be further examined in more detail in terms of explainability and interpretability, there may be some insights for creating a new type of artificial neurons. Inspired by the research, an equally powerful artificial neural network could be created without so many neurons.”

The paper Dendritic Action Potentials and Computation in Human Layer 2/3 Cortical Neurons is available on Science.

Journalist: Fangyu Cai | Editor: Michael Sarazen

We know you don’t want to miss any story. Subscribe to our popular Synced Global AI Weekly to get weekly AI updates.

Need a comprehensive review of the past, present and future of modern AI research development? Trends of AI Technology Development Report is out!

2018 Fortune Global 500 Public Company AI Adaptivity Report is out!
Purchase a Kindle-formatted report on Amazon.
Apply for Insight Partner Program to get a complimentary full PDF report.




We produce professional, authoritative, and thought-provoking content relating to artificial intelligence, machine intelligence, emerging technologies and industrial insights.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store


AI Technology & Industry Review — | Newsletter: | Share My Research | Twitter: @Synced_Global

More from Medium

Quantum Computing & AI

HPC-AI’s FastFold Shortens AlphaFold Training Time from 11 Days to 67 Hours

AI Governance

AI Needs Embodied Intelligence and Mortality