A groundbreaking revelation by researchers at the University of Cambridge, U.K., unveiled an artificially intelligent, self-regulating system that mirrors the human brain’s problem-solving approach. This remarkable development promises to usher in a new era of advancements in machine learning and neural networks while offering unprecedented insight into the mechanics of the human brain.
Like other intricate organs, human brains develop within certain limitations and conflicting needs. The need for optimization for information processing, balanced against the conservation of energy and resources, shapes the brain into an efficient system that operates within these physical restrictions.
Co-lead author of the study, Danyal Akarca, from the Medical Research Council Cognition and Brain Sciences Unit at Cambridge, notes, “Biological systems commonly evolve to maximize the use of available energetic resources. The results of this evolution are often elegant solutions that reflect the balance of competing forces.”
In collaboration with computational neuroscientist Jascha Achterberg, Akarca and his team crafted an artificial system that mimics a simplified version of the brain, operating under similar physical constraints. The resulting findings were published in the esteemed Nature Machine Intelligence journal on November 20.
Neurons, the fundamental units of the brain, form a complex network, creating information highways that span different brain regions. Replicating this structure, the team’s AI system used computation nodes assigned to specific locations in virtual space. Like neurons, the physical distance between these nodes determines the ease of communication. The system was then tested with a maze task requiring extensive information processing.
According to co-author Duncan Astle, this simple constraint of communication difficulty between distant nodes forced the artificial system to develop complex characteristics similar to those in biological systems like the human brain. This revelation underscores why our brains are structured the way they are.
Achterberg expressed his team’s astonishment at the results, noting that their AI system mirrors the brain in numerous ways. He highlighted two main areas in which their system resembled the brain: the internal structure, reflecting the intricate, energy-efficient connections of the brain, and the internal functions, mirroring the efficient signal transmission seen in the brain.
The team is optimistic that this AI system can be further refined to reveal how specific restrictions contribute to the unique characteristics observed in the human brain, particularly among individuals grappling with cognitive or mental health issues.
Co-author John Duncan believes these artificial brains could provide a novel way to understand the complex data collected from real neuron activity in actual brains.
In Achterberg’s words, “Understanding the brain’s problem-solving capabilities alongside its resource conservation goal can help us comprehend why brains are structured the way they are.”
