Despite having brains with the size of a sesame seed, bees can learn and recognize complex visual patterns with sophistication that rivals much larger animals. Scientists believe they have discovered the secret behind this impressive visual ability.
In a new study, recently published in, a team of researchers at Sheffield University in the United Kingdom neural model that mimics the way bees actively exploit their environment and process visual information.
The findings reveal how the bee brain employs a system of Standard recognition Surprisingly powerful and efficient that can revolutionize the way we build machines to see and think.
The model created by the team of researchers, led by neuroscientists and engineers, built a simplified but realistic model of the visual system of the bee, incorporating elements of neurobiologybehavior and automatic learning.
The study discovered how specific neurons in the insect brain, called “Lóbula neurons“They are self-organized in highly selective visual filters capable of coding detailed pattern information.
Even with Minimum neural resourcesthe model showed remarkable abilities, having successfully discriminated discriminating on mathematical symbols, which managed to generalize for new visual tasks, and even recognize human faces – all without learning by reinforcement or reward feedback, note o.
The fundamental discovery focuses on “active view“, A strategy where organisms do not passively receive visual input “Instead, they actively exploit their environment.”
Bees can do so through deliberate movements of the head and the flight, Building neural images over time instead of a single look.
“In this study, we successfully demonstrated that even the smallest brains can take advantage of the movement to perceive and understand the world around it,” says James MarshallDirector of the Artificial Intelligence Center of Sheffield University and main author of the study.
“This shows us that a small and efficient system can perform much more complex calculations than we thought we were possible,” adds the researcher.
One of the most notable findings of the study was the surprisingly small number of neurons needed for effective functioning. With only 36 Lóbula neurons, the model performed statistically significant in various recognition tasks. Even with only 16 neuronshe could recognize patterns such as spirals or inclined bars.
“The numbers are surprisingly small, even for complex tasks as recognition of human faces,” explains Lars ChittkaProfessor of Sensory and Behavioral Ecology at Queen Mary University of London and co-author of the study.
Discoveries may have various practical applicationswhich go beyond insect neuroscience.
Dependence on the model in simple and biologically founded learning rules opens doors for the development of neuromorphic systems for robotics, computer vision and autonomous navigation.
Instead of requiring massive neural networks with millions of labeled examples and vast computational resources, future AI systems can learn Like bees: actively collecting samples from the world, refining perception over time, and developing efficient internal codes.
