Recently, the artificial intelligence program ChatGPT has gone viral on the Internet because of its "erudite knowledge" and "answer questions". The generative AI represented by ChatGPT has strong content generation ability and the "intelligence" level of human beings, and its application in the military field is bound to have an impact on future wars.

Significantly improved battlefield awareness. In future wars, various new types of rapid killing weapons will further accelerate the rhythm of the battlefield, requiring continuous improvement of battlefield situational awareness capabilities, and thus supporting rapid decision-making to meet the needs of the battlefield. In the battlefield space full of "fog", in the face of massive multi-source, complex heterogeneous and rapidly growing battlefield situation data, human perception speed and processing ability appear to be "slow". The large visual model architecture introduced in recent years, through unsupervised pre-training and human feedback reinforcement learning paradigm, has made breakthroughs in many fields such as image classification, object detection, semantic segmentation, pose estimation, image editing and remote sensing image interpretation, which can significantly improve battlefield perception. Intelligent weapons embedded with large visual models can accurately identify and distinguish the primary and secondary, true and false targets through the visual system, quickly extract and generate high-value intelligence from massive multi-modal data, reduce the cognitive load of combatants, and form a comprehensive, timely and accurate judgment of the situation. Using the perceptual advantages of generative AI to achieve accurate positioning of key nodes may be the premise of launching combat operations in the future.

Greatly promote human-computer interaction. Human-computer interaction allows machines to "hear" human language, "see" human actions and expressions, "understand" human emotions and intentions, and present the calculation process and results in a way that is easy for humans to understand. The large language model can not only perform well in text understanding scenarios such as sentiment analysis, speech recognition, and information extraction, but also apply to the visualization generation scenarios of battlefield information systems such as picture description generation, book manuscript generation, and dialogue generation. If it is embedded in the integrated joint combat system and continues to evolve iteratively, it can be used for more complex work such as planning, operation plan generation, exercise results evaluation, and so on, and may reshape the command decision-making process in future wars. By deeply embedding the generative AI application of ChatGPT into the command information system, the intelligent equipment can "understand" the instructions, accurately understand and analyze the operational requirements of the commander through the man-machine dialogue between the commander and the battlefield information system, and generate action reference schemes on this basis, providing a new means for the rapid and reasonable allocation of combat forces in the future war.

Promote the autonomy of command decision-making. In information-based and intelligent war, the participating forces are diverse, the combat styles are diverse, and the battlefield situation is changing, and the commanders are facing the "bottleneck" of insufficient intelligence in effectively conducting the war. With the help of the intelligent auxiliary system of decision-making large model, the "man-machine" mixed decision-making mode may become a new choice. Although from the current level of technology, ChatGPT class generative AI applications are still unable to perform machine control, group collaboration, dynamic scheduling and other operations. However, its powerful parallel processing ability, which can handle thousands of tasks at the same time, is suitable for integrating human/unmanned platforms, generating control algorithms, optimizing group behavior, and fully supporting "swarm", "fish" and "Wolf pack" combat multi-agents. The command and control system based on decision large model can give full play to the advantages of both human brain and artificial intelligence, and realize the leap from intelligent prediction to intelligent decision, and from controlling single agent to multi-agent. In the future battlefield, the embedding of generative AI into the unmanned combat platform can innovate a new paradigm of military operations and greatly improve combat effectiveness.

A new model of logistics support has been created. From the perspective of scientific and technological development, military force confrontation is increasingly expanding to the full dimensions of the physical domain, the information domain and the cognitive domain, the combat space is extending to the extremely high, extremely far and extremely deep all-round, and the corresponding logistics support tasks are becoming more diversified and complex. In the future battlefield, the multi-mission general large-scale model will be integrated into the unmanned combat platform and various support systems, and people, equipment and objects will be interconnected, and various combat and support entities will be organically integrated. The logistics support system realizes intelligent matching between people and materials, materials and equipment, materials and troops, and materials and regions through deep learning analysis of big data such as the quantity, time, and maintenance of materials in storage, and automatically predicts material demand, matches better delivery vehicles, develops better transportation plans, and timely solves problems in the battlefield logistics supply link. Especially in the face of extreme combat support that is difficult for personnel to reach such as complex terrain, contaminated areas, and fire control areas, based on a large number of pre-training samples for special tasks, the generated AI can realize changes in demand perception, resource allocation, and action control, and independently assign tasks, plan paths, and navigate and position. The guarantee materials will be directly and accurately distributed to the guarantee objects in a "point-to-point" way to achieve intelligent guarantee.