Can AI Think Like Us? New Research Mimics Human Memory for Smarter Machines

A new AI framework inspired by human memory could make machines more efficient, adaptive, and capable of reasoning.

A recent paper published in the journal Engineering presents a novel approach to artificial intelligence by modeling it after how human memory functions. The research aims to overcome key limitations of current large-scale models like ChatGPT, setting the stage for more efficient and cognitively intelligent AI systems.

While large models have demonstrated impressive performance across a range of applications, they also exhibit significant shortcomings. These include high data and computational demands, susceptibility to catastrophic forgetting, and limited logical reasoning capabilities. According to the study, these issues arise from the fundamental design of artificial neural networks, their training processes, and their reliance on purely data-driven reasoning.

Introducing Machine Memory and the M2I Framework

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.

Four Focus Areas of the M²I Framework

The M²I framework centers around four key areas:

1. Neural Mechanisms of Machine Memory: The research investigates how neural systems in the brain are pre-configured and how brain development and plasticity contribute to intelligence.
2. Associative Representation: The framework aims to encode and retrieve information through associations such as abstract–concrete links and spatiotemporal connections, mimicking the way human memory organizes and retrieves knowledge.
3. Continual Learning: To tackle the issue of catastrophic forgetting, the researchers propose methods that support continual learning, even under low-power conditions. This enables AI systems to integrate new knowledge without losing previous information.
4. Collaborative Reasoning: The model aspires to combine intuitive and logical reasoning systems, enhancing both interpretability and efficiency in AI reasoning processes.

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.

Toward the Next Generation of Intelligent Machines

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.

Reference: “Machine Memory Intelligence: Inspired by Human Memory Mechanisms” by Qinghua Zheng, Huan Liu, Xiaoqing Zhang, Caixia Yan, Xiangyong Cao, Tieliang Gong, Yong-Jin Liu, Bin Shi, Zhen Peng, Xiaocen Fan, Ying Cai and Jun Liu, 28 January 2025, Engineering.
DOI: 10.1016/j.eng.2025.01.012

Never miss a breakthrough: Join the SciTechDaily newsletter.
Follow us on Google and Google News.