How far artificial neurons are from the real ones_post_6

Company

Company

How far artificial neurons are from the real ones

02

Neurons communicate through molecules. Each neuron finds its "target" not because it grows towards it or forms a synapse with it, but because the target "recognizes the signal", just as a cell phone would recognize and pick up its signal from the general field. The phone "knows" the signal is intended for it. There's no such thing as invariable, rigid neural networks.

The main actor in synapses is the mediator. The size of the largest neurons is less than 1/10th of a millimeter (the size of their body), while the axon (the outgrowth of the neuron that conducts the electrical signal) can be more than a meter long.

07

06

As long as the information remains inside the nerve cell, the transmission is electrical. Specific electrical impulses travel across the nerve cell membrane.

It seems that interactions between biological neurons CANNOT even be referred to as networks. This is some other method of interaction. Therefore, computer neural networks are a misinterpretation of neural communities.

03

For a network to form, an axon must reach the next cell and create a contact — a synapse. It is the synapse that is the elementary unit of brain function, not the neuron. A neuron is a complex computing machine that matches signals from thousands of channels — axons. The signal is transmitted between neurons in special synapses by chemical means.

05

04

It can be assumed that biological neural "communities" stand out from AI neural "networks" by different interaction methods, e.g., when the incoming elements have a NON-uniform structure, a community may have an inherent hierarchy, with its elements (members) being UNequal.

08

The human brain has up to 1 trillion synapses. A human synapse has between one thousand and 300,000 binary operators. Each synapse contains about 25 mediators, which can transmit signals differently by working together in different combinations. There are 100,000 to 1 million synapses per neuron.

But neurons are not the only brain cells that respond in one way or another to changes in the environment. A network of astrocytes generates signals, but, unlike neurons, astrocytes are not linked by synaptic connections. But how are "astrocytes" even leveraged in modern AI? Neurons, with their connections (albeit weighted), are at least somehow represented, while astrocytes (without connections) are nowhere to be seen.

09

Astrocytes, with their huge numbers and unclear functions, are another argument proving it difficult or even impossible to "extract" something from the brain and build a new, true AI using the toolkit currently available in both AI and science.

10

01

Neurons are known to have aspect specialization. Neurons have an accumulated effect and get their current specialization in the form of an accumulated (cumulative) total. This is not a quick process. Current understanding of neuronal functioning is not enough to explain all the cognitive specialization phenomena.