A new perspective on the future development of artificial intelligence (AI) is proposed by researchers Li Guo and Jinghai Li in an article entitled “Developing Artificial Intelligence: Towards the Consistency of Datasets, AI Models, Model Building, and Logical Structures of Hardware.” Engineeringで公開されています。 The authors argue that while current AI systems have made great strides in handling the statistical properties of complex systems, they face challenges in effectively handling and fully representing the spatial complexity patterns of these systems.

This paper begins by highlighting the global interest in AI and its potential applications, and the need for sustainable and long-term development. The authors question the current logical architecture of AI, particularly in the engineering domain, suggesting that ensuring consistency between the logical structures of datasets, AI models, model building software, and hardware could be an important direction for future AI development.

In engineering research, consistency between the logical structures of research objects, physical models, software systems, and hardware platforms is fundamental to ensuring the functionality, reliability, and scalability of application systems. Logical structures refer to the framework of a system, the basic building modules, and the interconnections and joint relationships between these modules. This consistency allows code to be easier to maintain, allowing it to replicate the evolution of space-time of research objects, and gives it a deeper understanding of its structural and functional properties.

The authors point out that current AI is based primarily on artificial neural networks (ANNs) and deep learning technologies, and is relatively rough and insufficient to reflect complexity principles. The success of AI applications in fields such as computer vision and natural language processing involves deep networks with trillions of parameters, but there is no logical relationship or structural correspondence between these parameters and the objects being modeled. This will turn the training process and model into a “black box” and disconnect from the spatio-temporal structural evolution patterns of complex systems.

This paper highlights that current AI logic architectures do not adequately reflect the multilevel, multiscale, and spatial properties unique to processing objects. AI models cannot implicitly extract, process, present, and utilize these physical properties in data in a reasonable, appropriate and effective way. This is a fundamental problem that needs to be addressed when applying AI to engineering systems.

The authors propose that the logic architecture of AI must evolve and better fit into the principles of multilevel complexity. They propose to integrate the Principles of Compromise (CIC) in multiple dominant mechanisms in the design, training, optimization, and application processes of AI systems. This principle, demonstrated in Mesoscience, can guide AI modeling and improve predictive capabilities.

To further explore the direction of AI development, the authors recommend several actions. First, we need to deepen our multi-level complexity principles to ascertain their necessity and rationality. Next, we need to select typical cases from the engineering domain to build datasets and AI models based on the new logical structure. Finally, a unified logical architecture must be established to define objects, AI models, model building software, and hardware with consistent logical structures.

The authors envision an engineering intelligentization paradigm based on the principles of multilevel complexity, including multilevel structures, multiscale structures at each level, three expert AI model buildings, and AI modeling of complex mesoregimes. This framework is expected to exhibit characteristics such as improved convergence stability and predictive performance, even for small training data sets.

This paper suggests that AI development must ensure consistent logical structures between datasets, AI models, software and hardware based on the principle of complexity. This approach could lead to the development of computational paradigms that integrate new AI systems, R&D models, and multilevel complexity science principles. The authors call for extensive interdisciplinary cooperation to address the challenges of incorporating physical principles into the logical architecture of AI.

Paper “Developing Artificial Intelligence: Towards the Consistency of Datasets, AI Models, Model Building, and Logical Structures in Hardware,” written by Li Guo and Jinghai Li. Full Open Access Paper: https://doi.org/10.1016/j.eng.2025.05.004