Japanese scientists have successfully developed a groundbreaking artificial intelligence system that models the brain using a method called “reservoir computing.” Through this innovation, researchers have utilized neurons from the rat brain’s cortex to create their own artificial “brain.” The study, which has been published in the journal Proceedings of the National Academy of Sciences, reveals that this system exhibits short-term memory capabilities and can effectively classify data sequences such as spoken numbers.
The human brain, an intricate network composed of billions of interconnected neurons, is responsible for processing information and enabling our perception of the world and bodily control. Despite extensive research, our understanding of the brain’s structure and neuronal properties and their impact on information processing remains incomplete.
Reservoir computing, a method that aims to mimic the brain’s function, involves employing a multitude of interconnected nodes to transform input data into a more complex form.
In this particular study, researchers utilized the toning calculation technique to construct an artificial “brain” using neurons from the cortex of rats. By leveraging optogenetics and fluorescence visualization of calcium, the scientists were able to monitor the responses of the artificial neural network. Subsequently, they processed the data using tankar calculation and discovered that the artificial brain possesses short-term memory capabilities, enabling it to effectively classify data sequences.
Furthermore, the artificial brain proved its ability to classify spoken numbers accurately even when the spoken prompts were varied. These findings demonstrate that biological neurons can function as filters to enhance reservoir computing.
This breakthrough research also showcased that a model trained on a specific dataset could successfully classify new sets of data within the same category. This revelation underscores the artificial brain’s capacity to filter information, thereby enhancing the overall effectiveness of reservoir computing.