A new perspective on the future development of artificial intelligence (AI) has been put forward by researchers Li Guo and Jinghai Li in their article titled “The Development of Artificial Intelligence: Toward Consistency in the Logical Structures of Datasets, AI Models, Model Building, and Hardware?” published in Engineering. The authors argue that while current AI systems have made significant strides in handling the statistical properties of complex systems, they still face challenges in effectively processing and fully representing the spatiotemporal complexity patterns of these systems.

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

In engineering research, consistency among 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. The logical structure refers to the system’s framework, its basic building modules, and the interconnections and collaborative relationships among these modules. This consistency ensures that the code is easy to maintain and capable of replicating the spatiotemporal evolution of the research object, allowing for a deeper understanding of its structural and functional characteristics.

The authors point out that current AI, primarily based on artificial neural networks (ANNs) and deep learning technology, is relatively rough and insufficient to reflect the principles of complexity. Successful AI applications in fields such as computer vision and natural language processing involve deep networks with trillions of parameters, but there is no logical relationship or structural correspondence between these parameters and the object being modeled. This results in the training process and model becoming "black boxes," disconnected from the complex system’s spatiotemporal structural evolution patterns.

The paper emphasizes that the current logical architecture of AI does not adequately reflect the multilevel, multiscale, and spatiotemporal characteristics inherent in the processing objects. AI models are unable to implicitly extract, process, present, and utilize these physical characteristics in the data in a reasonable, adequate, and effective manner. This is a fundamental issue that must be addressed when applying AI to engineering systems.

The authors propose that the logical architecture of AI should evolve to better align with the principles of multilevel complexity. They suggest integrating the principle of compromise-in-competition (CIC) among multiple dominant mechanisms into the design, training, optimization, and application processes of AI systems. This principle, which has been demonstrated in mesoscience, could guide AI modeling and improve its predictive capabilities.

To further explore the direction of AI development, the authors recommend several actions. First, the principles of multilevel complexity should be deeply investigated to confirm their necessity and reasonableness. Then, typical cases from the engineering domain should be chosen to construct datasets and AI models based on the new logical structure. Finally, a unified logical architecture should 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 multilevel complexity principles, which would include a multilevel structure, multiscale structure at each level, three-regime AI model building, and AI modeling in the complex mesoregime. This framework is expected to exhibit characteristics such as convergence stability and improved prediction performance, even with small training datasets.

The paper suggests that the development of AI should be guided by the principles of complexity science to ensure consistent logical structures among datasets, AI models, software, and hardware. This approach could lead to the development of a new AI system, R&D model, and computational paradigms that integrate the principles of multilevel complexity science. The authors call for extensive interdisciplinary cooperation to address the challenges of incorporating physical principles into the logical architecture of AI.

The paper “The Development of Artificial Intelligence: Toward Consistency in the Logical Structures of Datasets, AI Models, Model Building, and Hardware?” is authored by Li Guo, Jinghai Li. Full text of the open access paper: https://doi.org/10.1016/j.eng.2025.05.004. For more information about Engineering, visit the website at https://www.sciencedirect.com/journal/engineering.