What are dreams made of? The new scientific theory

A new theory explains why we have the "strangest" dreams, starting with Artificial Intelligence

There are many theories that have attempted, over the years, to give a scientific explanation to the fact that we dream. Some claim that dreaming is a simple "side effect" of the normal functioning of the brain, according to others instead, dreams have their own specific functionality.

The most popular theory about dreams, today, suggests that dreams are involved in the process of storing and preserving data collected during wakefulness.

Inspiration from Artificial Intelligence

A new theory - just published in the scientific journal Patterns by Erik Hoel of Tufts University in Medford, Massachusetts - proposes a new hypothesis that can explain dreams.

According to Hoel, dreams would be a tool that the human brain uses to generalize experiences. Inspired by the learning typical of neural networks, particular Artificial Intelligences that "simulate" the structures of the human brain, Hoel's theory is based on the assumption that our brain, like neural networks, needs to know chaos to be able to process reality.

The author calls the new theory the "brain hyper-adaptation hypothesis," and suggests that the nature of dreams might be related to a genuine defense mechanism against the brain's over-adaptation to the world as it presents itself during the waking state.

When a neural network is involved in a learning process, it is subjected to stimuli that resemble what it will have to look for once it has completed its "training," so-called machine learning.

It is common usage in machine learning to insert within the data sets submitted to the AI information intended to confuse, to randomize the stimuli so that the neural network does not become "narrow-minded," Hoel writes in the paper.

In practice, the chaotic contamination of stimuli is what allows AI to find patterns on a large scale, identifying an "overall pattern" that would otherwise be flattened on the individual data at hand.

Why do we have "weird" dreams?

According to new theory inspired by AI behavior, the irrational nature of dreams underlies our ability to work out an overall pattern of the world around us.

Come gli stimoli caotici aiutano le reti neurali, così i sogni aiuterebbero il cervello umano a non appiattirsi su quello che succede in stato di veglia ed essere così pronto a rispondere a pattern inaspettati o non previsti.

E sarebbe proprio la natura "allucinogena, senza regole e favolosa" dei sogni a permettere al nostro cervello di introdurre i "dati corrotti" sulla realtà tra gli stimoli da prendere in considerazione da svegli.
I compiti bizzarri e "fuori dal mondo" cui i sogni ci sottopongono sono un vero e proprio allenamento per il nostro cervello: inserendo dati randomici nel set a disposizione, i sogni permettono al nostro cervello di guardare al mondo in maniera più generale.

"Dreams are here to keep us from becoming over-adapted to the model of the world" as we know it during the day, Hoel writes, and thus be able to respond to unexpected stimuli and patterns.

The experimental evidence for this, according to the scientist, would already be known: the fact that when we perform a repetitive task during the day it is virtually certain that we will dream about it, would be a ready response of the brain to hyper-adaptation. And the strangest dreams would be an evolved self-defense mechanism put in place by the brain not to yield to the flattening on already known stimuli.