Decentralized physical infrastructure networks (DePIN) were an indispensable cornerstone for the upcoming digital society, transforming traditional infrastructure reliant on centralized control and management into a new type of infrastructure network driven by communities and autonomous governance. The deep integration of blockchain and artificial intelligence (AI) technologies played a key role in enabling this transformation. To this end, this paper first emphasized the organic combination of "AI for blockchain (AI4B)" and "blockchain for AI (B4AI)", forming a synergistic, bidirectional closed-loop circuit to construct the truly intelligent blockchains, thereby realizing a new paradigm of blockchain intelligence. Secondly, it proposed a novel technological architecture for intelligent blockchains: on one hand, embedding intelligence into every layer of traditional blockchain architecture to cultivate the blockchain intelligence ecosystem ranging from foundational intelligence to infrastructure intelligence and then to application intelligence; on the other hand, corresponding to the actual intelligent blockchain construction, artificial intelligent blockchain were built, and through their virtual-real feedback and parallel execution, the blockchain intelligence ecosystem evolved from static rule orientation to dynamic adaptive evolution. Then, it analyzed the core intelligent attributes of blockchain intelligence from the perspectives of intelligent contracts, decentralized identity, knowledge management, and autonomous governance. Lastly, it outlined the typical application scenarios of blockchain intelligence and the key challenges faced. This paper aims to actively promote the research and development of intelligent blockchains and blockchain intelligence technologies, laying the foundation for building the core infrastructure of the future digital age.

Embodied intelligence transcends the boundaries of traditional artificial intelligence by emphasizing the importance of interaction between machines and the physical world, facilitating the development of intelligent entities that combine hardware and software to learn from and adapt to their environments, thereby solving real-world problems. Inspired by this philosophy, the concept and framework of embodied intelligent driving are introduced, aiming at integrating the idea of embodied intelligence into the development and application of autonomous vehicles. Through the continuous interaction between physical agents, virtual agents, and real traffic scenes, intelligent driving systems can achieve precise perception, efficient execution, and autonomous evolution in complex scenes, enhancing the long-term adaptability of autonomous vehicles in open traffic environments. Based on the embodied intelligent driving framework, the relevant technologies are summarize and the development status and existing problems of such technologies are analyzed. Furthermore, thoughts and prospects in this field are demonstrated by exploring the important roles and application potential of virtual-real interactive data intelligence, foundation models and foundation intelligence, continuous learning and parallel intelligence. This paper is expected to promote innovative research and the application on embodied intelligent driving in a wider range of scenarios, and provide new ideas and solutions for the development of mobile robot systems such as intelligent vehicles.

This paper summarized the focus, the development trend and the technical nature of AI large model research, analyzed the development strategy of AI at the national level, the urgent needs in the field of national defense, and the basis of applications in the field of ships. Then, from the aspects of the development of intelligent green ships, the innovation of defense equipment systems, the construction of management and control system and the transformation of knowledge-intensive industries, the broad prospect of applying AI large model technologies to the field of ships was discussed. The paper pointed out that the combination of AI large model technologies and concepts such as parallel systems, knowledge factories and digital employees can catalyze new designs, research and development and verification methods such as "AI design" + "digital factory" + "parallel verification". In addition, AI large model technology can inject intelligent and green elements into the shipping industry from aspects such as the hull design, ship construction, shipping management, energy conservation and emission reduction, can optimize ship functions, and improve efficiency, economy and environmental protection. Combined with new materials, new energy power and new information electronics and other technologies, AI large model technologies can shape the future marine defense equipment system based on new concepts and new patterns. At the same time, AI large model technologies can enable the construction of ship management and control systems, optimize planning, help scientific and technological innovation, improve management efficiency and improve the quality and efficiency of the corporations. In particular, with the establishment of knowledge factories in the field of ships, the training of digital employees, the promotion of industrial robots and the expansion of far-reaching sea fields, artificial intelligence large model technologies will be able to promote the organic combination and close collaboration of "natural persons", "robots" and "digital people" in the field of ships, and accelerate the upgrading of the ship industry to be knowledge intensive and intelligent intensive. This can transform the industrial ecology and value creation mode to be high-end, intelligent, and green, and realize a development mode that pays more attention to quality and efficiency for shipbuilding corporations.

The advancement of large models offers new perspectives and foundation intelligence for building parallel human ecosystems comprised of biological humans, digital humans, and robotic humans. However, challenges such as time-limited updates to knowledge, inadequate specialized capabilities, and risks of information privacy leakage persist in the management and control of complex systems. To tackle these issues, a retrieval-augmented generation-driven parallel human and parallel intelligence framework (RAG-PHI) is introduced. It proposes to establish an open data platform that facilitates the integration of real-time, industry-specific, and private knowledge into the parallel human system. It develops dynamic routing and retrieval for context capture and the reconfiguration of parallel human capabilities, along with introducing context-aware prompt learning to enhance cognitive and behavioral skills. Furthermore, towards the organization and management, training and evaluation, operation and production of parallel human, the structures of parallel human community, parallel human school, and parallel human factory are proposed by the RAG-PHI architecture. These are designed to foster a parallel human ecosystem powered by RAG and large foundation models, thereby enhancing productivity in the age of intelligent industries.

In the wave of digitization, the introduction of digital and parallel intelligence technologies was crucial to responding to the imbalance in the allocation of medical resources and the need for optimization of hospital management. Parallel healthcare systems can fully leverage the value of data with the help of artificial intelligence methods and optimize the solution through the interaction between reality and virtuality. The system comprised the distributed and trustworthy database, medicine-oriented operating system, medicine-oriented scenario system, and medicine-oriented large models, which enhanced data empowerment through data collection and circulation, thus facilitating trans-dimensional optimization of hospital operations. To better demonstrate the value of data in hospital operations, a method to quantify the value of data was proposed based on parallel healthcare systems and demonstrated with the help of a numerical experiment. The experimental results showed that parallel healthcare systems can effectively improve the hospital's social benefit by 47.45%, economic benefit by 13.82%, and operational efficiency by 21.90%, providing a global vision for decision-makers.

Vertical federated learning (VFL) is commonly used for cross-domain data sharing in high-risk scenarios. Users need to understand and trust model decisions to promote the application of models. Existing research primarily focuses on the trade-off between explainability and privacy within VFL, and fails to fully meet the needs of users for establishing trust and fine-tuning models. To address these issues, we proposed an explainable vertical federated learning method based on human-in-the-loop (XVFL-HITL), which incorporated user feedback into the VFL's Shapley value-based explainability approach through a distributed HITL structure, using the knowledge of all VFL participants to correct training data and enhance model performance. Furthermore, considering privacy concerns, this paper employed the additive principle of Shapley values to integrate the feature contribution values of all entities other than the target participant into an aggregated measure, which effectively protected the feature privacy of each participant. Experimental results indicated that on benchmark data, the explainability results of XVFL-HITL were effective and could well protect the feature privacy of user. Additionally, compared to VFL-Random and VFL-Shapley, the model accuracy of XVFL-HITL improved by approximately 14% and 11%, respectively.

To address the low recognition accuracy of generic emotion recognition models when faced with different individuals, a multimodal individual emotion recognition technique based on joint labelling with integrated learning and clustering was proposed. The method first trained a generic emotion recognition model based on a public dataset, then anallysed the distributional differences between the data in the public dataset and the unlabelled data of individuals, and established a cross-domain model for predicting and labelling pseudo-labels of individual data. At the same time, the individual data were weighted clustered and labelled with cluster labels, and the cluster labels were used to jointly label with pseudo-labels, and high confidence samples were screened to further train the generic model to obtain a personalized emotion recognition model. Using this method to annotate these data with the experimentally collected data of 3 emotions from 3 subjects, the final optimized personalized model achieved an average recognition accuracy of more than 80% for the 3 emotions, which was at least a 35% improvement compared to the original generic model.

With the rapid development of the next generation of artificial intelligence technologies, such as the internet of things, large language models, and multimodal interactions, the traditional processes of drug research and production processing had been facing the challenges of an intelligent transition in recent years. In this context, this paper used the theory of parallel intelligence as the research philosophy and proposed a virtual-real interactive parallel drug systems, utilizing the ACP approach and large language models. It incorporated the concept of three types of beings—digital humans, robots, and natural persons—into the systems, providing a detailed discussion on the theoretical underpinnings, construction techniques, and potential application scenarios of the systems. The parallel drug systems covered the entire process of the pharmaceutical industry. For the drug development phase, it considered processes such as drug discovery, laboratory research, and clinical trials. In the production processing phase, it encompassed pharmaceutical manufacturing operations and system analysis predictions. The medical healthcare subsystem included personalized medication consultation, augmented reality drug guidance, and privacy security. The whole systems open up a digitized "drug space", aiming to establish a new paradigm for the drug systems and propel the revolution of intelligent medication.

Due to the potential issue of non-rigid deformations in the manufacturing process of mobile phone frames, most existing dispensing automation methods struggled to acquire accurate dispensing positions. For this purpose, an innovative approach of dispensing trajectory generation was proposed, which addressed the limitations associated with previous methods. Firstly, the phone frames were scanned using a 3D laser profiler, then a point cloud registration algorithm based on the pose graph was introduced to reconstruct the point clouds of frames. The template-based approach was employed, and the dispensing trajectory for template was manually annotated. Subsequently, the template was aligned with the target by a non-rigid registration algorithm based on the deformation graph, thereby the transformed trajectory was obtained. Furthermore, a trajectory refinement strategy was presented to generate the robust, accurate dispensing trajectory of the target. Extensive experiments demonstrate that the errors of the generated dispensing trajectories were consistently below 0.01 mm, and the speed can meet the real-time requirement of practical applications.