Chinese researchers develop a neural network to read raw sensory data

Beijing (ANTARA) – Chinese scientists have developed a new neural network that enables artificial intelligence (AI) to form concepts from raw sensory data such as vision and hearing, simulating fundamental aspects of human cognition, according to a study recently published in Nature Computational Science. One remarkable capability of the human brain is its ability to form abstract conceptual representations from sensorimotor experiences and apply them flexibly without relying on direct sensory input. However, the computational mechanisms underpinning this ability are not yet well understood. This implies that large language models are essentially constrained by their dependence on existing linguistic data, and thus are unable to spontaneously generate new concepts from experience-based learning. Researchers from the Institute of Automation under the Chinese Academy of Sciences (CAS) and Peking University proposed their new neural network framework, named CATS Net, as a way to address these limitations. The framework comprises a concept abstraction module and a task-solving module that can precisely instruct the framework to perform tasks such as recognition and evaluation while processing visual information, such as images. The framework can also independently generate a diverse set of new concepts, building its own unique “concept space.” Once the concept spaces from different AI systems are aligned, they can directly transfer knowledge using those concepts, without retraining on raw data. The process simulates how humans communicate using language. Through brain-imaging studies, the researchers revealed that the conceptual space built by CATS Net is highly aligned with human cognitive and linguistic logic, and its mode of operation is very similar to activity in the brain’s concept-processing areas. This shows the model does more than merely imitate brain function, while providing insights into the computational mechanisms humans use to form and employ concepts in the brain.