summary: Researchers have discovered that the basic building blocks of the human cortex are themselves powerful microchips. By using artificial intelligence to measure the computational complexity of a single cell, the researchers demonstrated that a single human cortical neuron is not a simple switch, but an extremely sophisticated computing device with processing power comparable to an entire multilayer deep artificial neural network (ANN).
important facts
- “Twin” imitation index: To mathematically define a neuron’s processing power, the team developed a brilliant new framework. They task sophisticated artificial neural networks with learning and perfectly replicating the input-output functions of a single human cortical brain cell. The more computational layers an artificial “twin” requires to mimic a living cell, the more complex a single neuron becomes.
- Advantages of dendritic trees: This study reveals that human cortical neurons have a significant computational advantage over other mammals due to their extremely dense and abundantly branched dendrites and highly specialized and unique electrical properties.
- Single cell pattern recognition: Rather than requiring a huge web of thousands of cells to perform basic identification tasks, a rich branching structure allows single Human cortical neurons independently perform sophisticated computations, such as processing visual input and distinguishing between completely different complex images (e.g., cat vs. dog).
- Challenge to scale theory: These findings directly challenge the long-held scientific theory that human intelligence is simply a product of brain size and total number of cells. Instead, human evolution prioritized the depth of internal calculations within the individual components themselves.
- New templates for neuromorphic AI: Today’s state-of-the-art machine learning models are built from highly simplified, unified mathematical points. This research provides a general framework that links physical cell shape to raw processing power, providing a blueprint for a revolutionary new generation of brain-inspired AI built from artificial units that are deep and multilayered in nature.
sauce: Hebrew University of Jerusalem
Why is the human brain capable of language, imagination, mathematics, and invention?
For many years, the prevailing view was that the secret to human intelligence lies primarily in scale: the enormous number of neurons in the human brain (nearly 100 billion) and the vast network of connections between them. But new research suggests that part of the answer may lie on a much smaller scale: the extraordinary computational power of individual brain cells.
Researchers have discovered that neurons in the human cortex are much more complex information processing units (“microchips”) than neurons in other mammals. The findings suggest that components of the human cerebral cortex may themselves be powerful in their own way, offering a possible explanation for how humans developed such extraordinary cognitive abilities.
The study was led by researchers and professors at Hebrew University. Idan Segev and Micky London, along with PhD students Ido Eisenbad and Daniela Yoeli, conducted the research at the Edmond Lilly Safra Brain Science Center (ELSC) in collaboration with Professor Chris de Kock from the Vrije Universiteit Amsterdam.
“People often think of neurons as simple switches that turn them on or off,” Segev says. “What we’re showing is that a single human neuron is itself a very sophisticated computing device.”
To make this discovery, the researchers developed a new method to measure the computational complexity of individual neurons. They used advanced computer models and artificial intelligence to assess how difficult it is for state-of-the-art artificial neural networks (ANNs) to learn and reproduce the input-output behavior of a single brain cell.
The more difficult it is for a “twin” artificial network to mimic a neuron’s input/output functions, the more computationally powerful that neuron will be.
This result indicates that human cortical neurons have a significant advantage in computational ability. Thanks to their richly branched dendritic trees and unique electrical properties, these cells can perform surprisingly complex calculations on input information, such as visual input (for example, distinguishing between images of cats and dogs).
This means that a single human cortical neuron is not just an “on/off” component in the brain. This is already an advanced computing unit in its own right, with computational power comparable to that of a deep neural network.
This finding challenges the traditional view that intelligence emerges primarily from the number of neurons and the connections between them. Rather, they suggest that the sophistication of neurons themselves may have played an important role in the evolution of human cognition.
The study also provides a new systematic and general framework for linking the physical structure of brain cells to their computational power, allowing scientists to move one step closer to understanding how the human brain produces thinking, learning, and cognition.
This research could also inspire a new generation of brain-inspired AI. This AI will be built from artificial units that are themselves computationally powerful and powerful, more similar to biological neurons, and very different from the highly simplified units underlying today’s most advanced machine learning systems.
Answers to key questions:
answer: That means current estimates of the brain’s computational capacity are vastly underestimated. For decades, computer scientists have likened the human brain to a digital network by treating each neuron like a single transition transistor or switch. If a single human neuron actually has the mathematical processing depth of an entire multilayer deep artificial network, the human brain essentially becomes a giant network. made of networkwith astronomical levels of computing power that completely dwarfs today’s supercomputers.
answer: The secret lies in the cell’s magnificent physical structure, especially its richly branched dendritic structure. Human cortical neurons do more than passively receive incoming electrical signals. Its large and expansive branch structure acts like a series of specialized sub-processors. These dense branches, combined with the unique membrane electrical properties, allow a single cell to simultaneously perform complex nonlinear calculations on multiple streams of incoming data before deciding to fire.
answer: While impressive, today’s state-of-the-art AI systems are built from billions of super-simplified, uniform mathematical points, making them highly inefficient and requiring large server farms. Professor Idan Segev says the research provides a concrete blueprint for a new generation of “brain-inspired” AI. By replacing these basic flat processing points with artificial nodes that mimic the deep, multilayered computational power of biological human neurons, we can build extremely powerful, compact, and energy-efficient AI networks.
Editorial note:
- This article was edited by the editors of Neuroscience News.
- Journal articles were reviewed in full text.
- Additional context added by staff.
About this neuroscience and cognition research news
author: Danae Marx
sauce: Hebrew University of Jerusalem
contact: Danae Marx – Hebrew University of Jerusalem
image: Image credited to Neuroscience News
Original research: Open access.
“Dendrite morphology and synaptic nonlinearity enhance the functional complexity of human cortical neurons,” by Daniela Yoeli, David Beniaguev, Idan Segev, and Ido Eisenbad. PNAS
DOI:10.1073/pnas.2533168123
abstract
Dendritic morphology and synaptic nonlinearity enhance the functional complexity of human cortical neurons
Humans exhibit unique cognitive abilities within the animal kingdom, but the neural mechanisms driving these advanced abilities are still poorly understood.
Human cortical neurons differ from neurons in other species, such as rodents, in both morphological and physiological features. Can the unique properties of human cortical neurons help explain our superior cognitive abilities?
Understanding this relationship requires measures that quantify how neuron properties contribute to the functional complexity of a single neuron. However, no such standardized scale currently exists.
Here, we propose the functional complexity index (FCI), a general deep learning-based framework for assessing the complexity of neuronal input and output. By comparing FCI of cortical pyramidal neurons across layers in rats and humans, we identified important morphoelectrical factors underlying the functional complexity of neurons.
Human cortical pyramidal neurons are significantly more complex functionally than rat pyramidal neurons, primarily due to differences in dendritic membrane area and branching patterns, as well as in the density and nonlinearity of NMDA-mediated synaptic receptors.
These findings reveal the structural and biophysical basis of the enhanced functional properties of human cortical neurons and provide an important step toward understanding the basis of enhanced cognitive abilities.
