Brain-mimicking AI reveals origins of biological intelligence
According to this article, researchers at the University of Cambridge in the United Kingdom have developed a self-organizing artificially intelligent system that solves particular problems by employing the same approaches as the human brain.
This research may offer new insights into the inner workings of the human brain in addition to helping the development of more effective neural networks in the field of machine learning.
A set of limitations and conflicting needs shape the development of the human brain and other complex organs. For instance, we need to improve our neural networks to process information efficiently while consuming minimal energy and resources. Our brains are shaped by these trade-offs to produce an effective system that works within these physical limitations.
“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. “The solutions they come to are often very elegant and reflect the trade-offs between various forces imposed on them”.
In order to represent a simplified version of the brain, Akarca and his team built an artificial system with imposed physical constraints in collaboration with co-lead author and computational neuroscientist Jascha Achterberg from the same department. The journal Nature Machine Intelligence reported their findings.
Neurons, which are connected brain cells, form the intricate network that makes up our brains. Together, these neurons create information highways that connect various parts of the brain. The team’s artificial intelligence system used compute nodes, each assigned a specific location in virtual space, in place of real neurons. Additionally, just like human brains, communication between two nodes became more difficult the farther apart they were. After that, the system was given a maze task to complete, which required processing information and many inputs.
“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. “Interestingly, they are characteristics shared by biological systems like the human brain. I think that tells us something fundamental about why our brains are organized the way they are”.
Put another way, the system started to employ some of the same strategies that real human brains employ to accomplish this particular task when it was subjected to physical constraints comparable to those that affect the human brain.
“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 into two groups”:
- The internal structure of the AI system resembles that of the human brain. This indicates that the connections between the different parts and neurons of the AI are comparable to those between the various regions of the human brain. In particular, the AI system exhibits extremely “brain-like” and energy-efficient internal wiring.
- The internal functions of the AI system resemble those of the human brain as well. This indicates that the signals produced by neurons to transmit data through the AI system’s connections resemble the signals found in the brain quite a bit. Once more, impulses from the brain are thought to be a particularly effective means of conveying information.
The goal of the team’s AI system is to provide light on the ways in which specific constraints contribute to the variations observed in the human brain, especially in individuals who experience problems related to cognitive or mental health.
“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 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 research could also help create AI systems that are more effective, especially for those who need to analyze a lot of data that is constantly changing while using a limited amount of energy.
“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 toward 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”.
These recent findings suggest to us how some technological aspects tend to come closer and closer to biological aspects. In this sense, we may assume that we are a ‘mere’ technological evolution at the highest level.