Scientists at the University of Cambridge in the U.K. have created a self-organizing, artificially intelligent system that uses the same tricks as the human brain to solve specific tasks.

This discovery not only allows for the development of more efficient neural networks within the field of machine learning, but may also provide new insights into the inner workings of the human brain itself. One of the study authors told Newsweek they were "very surprised" by the results.

The human brain and other complex organs develop under a set of restrictions and competing demands. For example, our neural networks must be optimized for information processing but not use up too much energy or resources. These trade-offs shape our brains to create an efficient system that functions within these physical constraints.

"Biological systems commonly evolve to make the most of what energetic resources they have available to them," co-lead author Danyal Akarca, from the Medical Research Council Cognition and Brain Sciences Unit at the University of Cambridge, said in a statement. "The solutions they come to are often very elegant and reflect the trade-offs between various forces imposed on them."

Together with co-lead author Jascha Achterberg, a computational neuroscientist from the same department, Akarca and his team created an artificial system with imposed physical constraints intended to model a simplified version of the brain. Their results were published in the journal Nature Machine Intelligence on November 20.

Our brains are made up of a complex web of interconnected brain cells called neurons. These neurons link together to form information highways that spread across different areas of the brain. Instead of using real neurons, the team's AI system used computation nodes, each given a specific location in virtual space. And, like our neurons, the further away two nodes were, the harder it was for them to communicate. The system was then asked to perform a maze task requiring multiple inputs and information processing to solve.

"This simple constraint—it's harder to wire nodes that are far apart—forces artificial systems to produce some quite complicated characteristics," co-author Duncan Astle, a professor from Cambridge's Department of Psychiatry, said in a statement. "Interestingly, they are characteristics shared by biological systems like the human brain. I think that tells us something fundamental about why our brains are organised the way they are."

In other words, when the system was put under similar physical constraints to those applied to the human brain, it began to use some of the same tricks used by real human brains to solve this specific task.

"We were very surprised by our results," Achterberg told Newsweek. "The AI system that we create in our work is similar to the brain in many ways. The many features we describe in our paper can roughly be grouped in two groups:

1. The AI system shows an internal structure similar to the human brain. That means that the ways individual parts and neurons of the AI are connected is similar to the way that different parts in the human brain are connected. The AI system specifically shows a very 'brain-like' and energy efficient internal wiring.
2. The AI system also shows internal function similar to the human brain. That means that the signals created by neurons to send information across the connections of the AI system look very similar to the signals we observe in the brain. Again, the brain's signals are thought to be a very efficient way of sending information."

The team hope that their AI system could be developed to shed light on how specific constraints contribute to the differences we see in the human brain, particularly those seen in people struggling with cognitive or mental health difficulties.

"These artificial brains give us a way to understand the rich and bewildering data we see when the activity of real neurons is recorded in real brains," co-author John Duncan said.

Achterberg said: "We show that considering the brain's problem solving abilities alongside its goal of spending as few resources as possible can help us understand why brains look like they do.

"Artificial 'brains' allow us to ask questions that it would be impossible to look at in an actual biological system. We can train the system to perform tasks and then play around experimentally with the constraints we impose, to see if it begins to look more like the brains of particular individuals.

"[Our research] strongly suggests that while the brain has all these very complex characteristics and features that we observe across studies within neuroscience, there might be very simple underlying principles causing all these complex characteristics."

Their findings may also contribute to the development of more efficient AI systems, particularly those that have to process large amounts of constantly changing information with limited energetic resources.

"AI researchers are constantly trying to work out how to make complex, neural systems that can encode and perform in a flexible way that is efficient," Akarca said. "To achieve this, we think that neurobiology will give us a lot of inspiration. For example, the overall wiring cost of the system we've created is much lower than you would find in a typical AI system."

Achterberg said: "Brains of robots that are deployed in the real physical world are probably going to look more like our brains because they might face the same challenges as us. They need to constantly process new information coming in through their sensors while controlling their bodies to move through space towards a goal. Many systems will need to run all their computations with a limited supply of electric energy and so, to balance these energetic constraints with the amount of information it needs to process, it will probably need a brain structure similar to ours."

Owning a digital twin of your brain is one step closer to becoming a reality

 

 

Recent breakthroughs in neuroscience and AI (Artificial Intelligence) have spurred innovative ventures into better understanding the intricacies of intelligence. Spearheading these efforts is a groundbreaking new platform called the Digital Twin Brain.

Developed by a team led by Tianzi Jiang at the Institute of Automation of the Chinese Academy of Sciences, this platform has the potential to revolutionize our understanding of both biological and artificial intelligence.

Understanding the Digital Twin Brain

The concept

Drawing parallels between biological and artificial networks, the Digital Twin Brain aims to create a digital replica of the human brain. This "twin" could absorb insights from biological intelligence, thereby offering a platform to bridge the two forms of intelligence.

Ultimate goals of Digital Twin Brain

The overarching objectives of the Digital Twin Brain are twofold:

1. Propel the evolution of Artificial General Intelligence (AGI).
2. Enhance the precision in mental healthcare.

These ambitions underscore the need for collaboration across multiple scientific disciplines on a global scale.

Potential applications

The Digital Twin Brain offers a sandbox for simulating various states of the human brain. Researchers can understand regular brain functions, pinpoint malfunctions linked to disorders, and devise techniques to alter undesirable brain states.

Core pillars of the Digital Twin Brain

Brain atlases

Brain atlases are akin to structural blueprints, ensuring that the digital replica mimics the intricacy of the biological brain. They encompass:

These atlases are imperative for a comprehensive understanding of brain regions, their interconnections, and the overarching organizational principles of the brain.

However, they also present challenges. Ensuring "biological plausibility" means that neural models are bound by the confines of these atlases, creating technical complexities.

Multi-level neural models

Derived from biological data, these models simulate brain functions. Their effectiveness relies heavily on a dynamic brain atlas, which, when enhanced, leads to more accurate function simulations.

Practical applications

The "twin", made up of the neural models, undergoes validation across diverse practical applications, such as disease biomarker identification and drug testing. The insights from these applications cyclically refine the brain atlas, ensuring an evolving and dynamic platform.

The Brainnetome Atlas

An essential asset in the development of the Digital Twin Brain is the Brainnetome Atlas. Introduced by the Institute of Automation of the Chinese Academy of Sciences in 2016, this atlas provides a macroscale overview of 246 brain sub-regions. It aims to offer an exhaustive and intricate mapping of the structure and connectivity of the human brain.

Need for a comprehensive platform

The current brain simulation platforms, despite their potential, lack a robust anatomical foundation. It's crucial to craft an open-source, versatile platform. This platform should be adaptable, user-friendly, and potent enough to endorse multiscale and multimodal modeling.

Open challenges of Digital Twin Brain

Several challenges loom over the Digital Twin Brain:

In summary, the Digital Twin Brain emerges as a nexus of neuroscience and AI. Integrating detailed brain atlases, evolving neural models, and varied applications, this platform stands on the cusp of reshaping our comprehension of both biological and artificial intelligence.

With the unified efforts of the global scientific community, it promises groundbreaking insights into the human psyche, advanced intelligent technologies, and novel treatments for brain-related conditions.

The full study was published in Intelligent Computing, a Science Partner Journal.