When we talk about artificial intelligence (AI), we often imagine systems that can analyze and process huge amounts of data, making predictions and decisions quickly. However, the capabilities of AI today are still far from natural intelligence – the spontaneous and flexible abilities we see in many animals in the wild. In a recent study, scientists Angelo Forli and Michael M. Yartsev from the University of California, Berkeley, took a deep look at the neural basis of natural intelligence and asked an important question: are we on the right track in studying the nervous system, or do we need a different approach to better understand natural intelligence?
According to Forli and Yartsev, natural intelligence is not just a fixed set of neural responses, but a self-organizing system that continuously adapts to changes in the environment. This leads to a new line of research that goes beyond analyzing individual parts of the nervous system to studying how the entire system works together to produce adaptive and flexible behaviors.
“Reductionism”
Forli and Yartsev’s report points out that much of current neuroscience research relies on reductionism—the practice of breaking complex systems into smaller, easier-to-study components. For example, scientists might look at how a group of neurons responds when a rat presses a button to receive food, or analyze the brain region responsible for controlling an animal’s basic movements. While this approach makes it easy to analyze individual parts of the nervous system, the authors say it can overlook the complex interactions between its components and miss important self-organization mechanisms of natural intelligence.
Forli and Yartsev argue that by separating out the components, we may lose the overall picture of how neurons and brain regions work together. They give a concrete example: when a pigeon returns to its nest, it relies not just on a few groups of neurons but on its entire nervous system working together to process information about its location, direction, and surroundings. This suggests that naturalistic intelligence is an integration of many components working in concert, far beyond a single set of responses. So the authors propose that, in order to understand naturalistic intelligence, a shift in research methodology is needed: from analyzing its parts to taking a more holistic approach.
Natural and laboratory environments
The two scientists also emphasized the difference between animal behavior in the wild and in the laboratory. In fact, most current neuroscience research is conducted under tightly controlled conditions, with simple, repetitive tasks. A mouse might have to run through a maze or press a button repeatedly to get a reward. While this makes it easier to control and measure the responses of neurons, it lacks naturalness and can miss important aspects of animal behavior. In the wild, animals are faced not only with simple tasks but also with unpredictable situations that require complex and creative responses, such as avoiding predators or finding food in harsh conditions. The authors argue that only by studying animal behavior in environments that are closer to nature can we truly understand how natural intelligence works.
Because biology is “diverse”
To understand naturalistic intelligence, we should not limit our research to a few laboratory animals, such as mice or monkeys. In nature, there are species that are uniquely suited to their environment, such as the precise navigation of pigeons or the ability of some amphibians to regenerate their own neurons. Each species has a unique nervous system, adapted to its environment and its way of surviving. By expanding the study to a wider range of species, scientists can discover new principles about how natural nervous systems process information and adapt to complex environmental changes.
Technology opens new horizons
Applying advanced tools helps scientists push the boundaries of modern neuroscience research. Technologies such as animal-mounted sensors, portable measurement devices, and high-resolution cell imaging allow scientists to collect more precise data from animals in their natural environments.
For example, head-mounted cameras can track behavior, while motion and positioning sensors can record an animal’s location and activity without interrupting or affecting its behavior. Advanced simulation tools, combined with artificial intelligence, are helping scientists more accurately recreate natural behavior in the lab. These advances open up new avenues for studying natural intelligence that have previously been difficult to study.
Conclude
In Forli and Yartsev’s view, understanding natural intelligence is not only a scientific goal, but also a way to expand our knowledge of what makes a flexible and adaptive intelligent system. Relying on reductionism alone will cause us to miss important elements of natural intelligence. To overcome this, we need to change our approach, from focusing on single experiments in the laboratory to observing animal behavior in the natural environment, as well as studying many different species.
By applying the principles that nature has developed over millions of years, we can develop artificial intelligence systems that go beyond current limitations. An AI based on natural intelligence principles would not be a dry data processing machine but could learn and creatively adapt to the environment, opening up new possibilities for technology and for human understanding of ourselves.
According to Understanding the neural basis of natural intelligence – Angelo Forli and Michael M. Yartsev, University of California, Berkeley. https://doi.org/10.1016/j.cell.2024.07.049 Cell.com Magazine