A recent paper published in Engineering titled “Machine Memory Intelligence: Inspired by Human Memory Mechanisms” explores a novel approach to AIby drawing inspiration from the human brain’s memory mechanisms. This research aims to address the limitations of current large models, such as ChatGPT, and paves the way for the development of more efficient and intelligent machines.

Large models have achieved remarkable performance in various fields but suffer from several drawbacks. They consume excessive amounts of data and computing power, are prone to catastrophic forgetting, and lack logical reasoning capabilities. These limitations stem from the architecture of artificial neural networks, training mechanisms, and data-driven reasoning methods.

To overcome these challenges, the researchers propose the concept of “machine memory,” a multi-layered, distributed network storage structure that encodes external information into a machine-readable and computable format. This structure supports dynamic updates, spatiotemporal associations, and fuzzy hash access. Based on machine memory, they introduce the M²I framework, which consists of representation, learning, and reasoning modules, forming two interactive loops.

The M²I framework focuses on four key areas. First, it explores the neural mechanisms that support machine memory, including how neurons and neural networks are pre-configured in the brain. Second, it aims to achieve associative representation in machine memory, such as abstract–concrete and spatiotemporal associations. Third, it focuses on continual learning under low-power conditions, addressing the issue of catastrophic forgetting. Fourth, it seeks to realize the dual-system cooperation of intuition and logic in reasoning.

In each of these areas, the researchers review the key issues and recent progress. For example, in the neural mechanisms of machine memory, they discuss how the brain’s development and plasticity contribute to intelligence. In associative representations, they explore ways to improve the encoding and retrieval of information in machine memory. In continual learning, they propose methods to adapt to new knowledge without forgetting old information. And in collaborative reasoning, they aim to enhance the interpretability and efficiency of reasoning in AI systems.

This research has the potential to revolutionize the field of AI. By mimicking the human brain’s memory mechanisms, the M²I framework could lead to the development of more intelligent and efficient machines that can better handle complex tasks and adapt to changing environments. However, further research is needed to fully realize the potential of this approach.

The study of machine memory intelligence inspired by human memory mechanisms offers a promising new direction for AI development. It provides a fresh perspective on addressing the limitations of current large models and has the potential to drive the next generation of intelligent machines. As the research progresses, it will be interesting to see how these ideas are translated into practical applications and how they impact various industries.

The paper “Machine Memory Intelligence: Inspired by Human Memory Mechanisms,” authored by Qinghua Zheng, Huan Liu, Xiaoqing Zhang, Caixia Yan, Xiangyong Cao, Tieliang Gong, Yong-Jin Liu, Bin Shi, Zhen Peng, Xiaocen Fan, Ying Cai, Jun Liu. Full text of the open access paper: https://doi.org/10.1016/j.eng.2025.01.012.