Scientists have developed a neurocomputer model to create artificial intelligence

(ORDO NEWS) — The new model of the human brain consists of three to four levels and covers all cognitive functions. Researchers believe that it can be used to find the causes of mental illness and endow neural networks with artificial consciousness.

An international group of biologists from Canada and France has created a neurocomputational description of the mechanisms underlying the cognitive and mental activity of the brain.

As the researchers write in Proceedings of the National Academy of Sciences, this model is sufficient to fill critical gaps in our understanding of AI and the biological mechanisms of mental disorders.

And in the future, it will ensure the transition from underdeveloped artificial intelligence to full-fledged machine consciousness.

The model consists of three levels. The first is sensorimotor: it describes how the internal processes of the brain identify patterns in the flow of sensory data and associate them with certain movements.

At the second, cognitive level, the brain combines these patterns with each other, forming single and consistent contexts.

The third, conscious level of the model explains how the brain separates itself from the outside world and manipulates past patterns that exist only in memory.

The tasks that are performed at these levels increase from simple to complex. For each of them, the brain uses a separate fundamental mechanism.

First, incoming data is subjected to local unconscious processing by Hebbian learning . Its essence can be expressed as follows: neurons that are activated together from time to time are connected to each other by processes.

At the next, cognitive level, the processed data is already integrated on a global (for the brain) scale, but still without the participation of consciousness.

This happens through reinforcement learning in the form of a dopamine reward, thanks to which even distant neurons can connect with each other.

At the last level, consciousness comes into play. To model it, biologists used the neural global workspace theory. It consists of neurons with long axons (transmitting processes) that form interconnected tracts.

By selectively exciting and inhibiting separate zones of the previous two levels, these neurons ensure the functioning of memory, activating the necessary neural ensembles without the participation of external stimuli, which is why the recalled images “fly before the eyes.”

Also, neural tracts organize a two-way exchange of information between individual zones. Thanks to this, a person can comprehensively comprehend the phenomena of the external world and arbitrarily connect them with each other, even if this connection is not obvious.

All these processes together create an integral experience of subjective experience. It consists of a “hanging” sensory image, which is fed by memory and corrected by new incoming information.

Associate professor of computational psychiatry and one of the authors of the work, Guillaume Dumas , says that the new model may be enough to create the next generation of artificial intelligence.

Or even a full-fledged machine mind, if the neural global workspace theory is correct. To speed up the process, you can add a fourth, social level, at which individual intellects will begin to interact with each other.

The team made its first attempts to model this level ten years ago. It is social interaction that scientists consider to be the “dark matter” of AI, which makes the human mind such a powerful and versatile tool.


Contact us: [email protected]

Our Standards, Terms of Use: Standard Terms And Conditions.