Artificial Brains Need Sleep Equal to Regular Brains

The combination of software and hardware with cognitive abilities similar to the brain of the animal or human is that we call an artificial mind.

Sleep states can help neural networks to regain net stability after continuous self-learning cycles.

We don’t know if androids dream of electric sheep, but they will almost probably need rest periods that offer the same benefits to those that sleep offers to living brains.

However, a new study by scientists at the Los Alamos National Laboratory points that androids need rest time to function at their best, just like our human brain needs sleep.



All kinds of brains need to sleep.

All kinds of brains need to sleep.

Sleep is incredibly useful and essential for humans and animals to function properly. So soon the artificial brains will require similar requirements.

Los Alamos National Laboratory computer scientist Yijing Watkins explained the motivation of the research team for the study:

We are fascinated by the idea of training a neuromorphic processor in a way similar to how humans learn from their environment during child growth. 

Like our brains, Watkins and his team have observed that neural simulations become unstable after a long period of restless self-study. And when the team put these simulations into sleep states, stability was restored.

All kinds of brains need to sleep.

Garrett Kenyon, the study’s co-author and computer scientist at Los Alamos, said that the trickiest part of the research was finding a way to prevent neural networks from becoming unstable.

The idea of how to stop learning systems from becoming unstable only really arises when you try to use biologically realistic and enriched neuromorphic processors or when you try to understand biology itself.

The vast majority of machine learning, deep learning, and artificial intelligence researchers never encounter this problem because in the highly artificial systems they study, they have the luxury of performing global mathematical operations which have the effect of regulating the overall dynamic gain of the system. 

The team’s last resort trying to maintain network stability was to figure out how to simulate a sleep state for the artificial brain. The noise was the answer.

Creating a noise similar to the static you hear when tuning in to a radio station did the trick. The best option like something called Gaussian noise, which involves a wide range of frequencies and amplitudes.

According to research, this type of noise helps stabilize neural networks and does not hallucinate because it provides a much-needed resting time.