Scientists have long dreamed of building computers as powerful as the human brain.
Now the task has become much more difficult.
Not because the development of computers has encountered a bottleneck, but because the human brain is much more complicated than previously thought.
Spencer Smith and his colleagues at British and American universities found that dendrites, the smaller parts of neurons, are also active components that do their own "computations".
Until now, scientific knowledge established that only axons, the largest portions of neurons, would be active, and all brain processes would result from the action of neural networks, sets of several neurons firing their axons in a coordinated manner.
In this interpretation, dendrites are just the "wiring" that connects neurons to each other.
"Suddenly, it's as if the brain's processing power has become much greater than we originally thought", comments Smith.
"Imagine that you reverse-engineered a piece of alien technology, and what you thought was a simple wiring was actually transistors that compute information. It was something we just discovered", compares the researcher.
Computational power of the brain
The brain's greatest computational power was revealed when the researchers developed a technology with sufficient resolution to measure the electrical activity of the dendrites, which revealed its ability to generate its own electrical firing independently.
It was a similar technology that allowed, less than two weeks ago, researchers to be able to measure the electrical current of an individual synapse for the first time.
The experiment consisted of inserting a microscopic glass pipette, filled with saline solution, into neuronal dendrites in the brain of a mouse. This allowed to "hear" the signaling process directly from each dendrite.
The data collected proved that the dendrites effectively act as true "sub-neural computers", actively processing neuronal signals on their own, without depending on axons and even less on neural networks.
Thus, the estimates that are always made about the amount of neurons in the brain – the latest calculations indicate 86 billion – no longer serve as a parameter for the computational power of the brain because there are calculations being carried out internally in the neurons, in an amount and speed that still will need to be calculated.
Currently, attempts to mimic brain cognitive processes in computers are focused on so-called neuromorphic processors, usually built not with transistors, but with a memory-capable electronic component called a memristor.