The complexity of a single brain neuron
The processing of information that must somehow take place in our brain obviously involves great complexity. This complexity has generally been seen as largely arising from the ways in which vast numbers of neurons (nerve cells) can connect to and interact with each other. It is increasingly being recognized that for many key types of brain nerve cells, the individual neurons themselves are “highly complicated I/O [input/output] information processing devices.”
Beniaguev et al. utilized advances in machine (computer) learning, namely deep artificial neural networks that mimic how our neurons communicate with each other, to mimic the I/O operations of a ‘biological’ neuron model. This included its elaborate architecture and other properties, as well as the “large number of excitatory and inhibitory inputs that bombard the neuron.”
As reported by Whitten, the study found that to represent the computational complexity of a single such neuron model required “about 1,000 artificial neurons for just one biological neuron.” Whitten mentioned computational neuroscientist Grace Lindsay as cautioning the study was “only comparing a model to a model.” It was currently impossible “to record the full input-output function of a real neuron”. The model of the neuron was likely not capturing everything. “In other words, real neurons might be even more complex.”
There are roughly 86 billion neurons in our brain, each interacting with others in sophisticated networks. And like all cells, neurons are a world of structural and biochemical complexity in and of themselves. It’s hard to argue with the oft-cited phrase that the human brain is the most complex organization of matter in the known universe. Yet it’s widely taught that it evolved, with no creative design, plan, or purpose.
- Whitten, A. How computationally complex is a single neuron? quantamagazine.org, 2 Sep 2021.
- Beniaguev, D. et al., Single cortical neurons as deep artificial neural networks, Neuron 109(17):2727–2739, 1 Sep 2021.