Current Status, Patterns, and Prospects of AI Large Model Industrial Applications

Current Status, Patterns, and Prospects of AI Large Model Industrial Applications

by

Click the Blue Text to Follow Us

Current Status, Patterns, and Prospects of AI Large Model Industrial Applications

Source | Advanced Manufacturing

1. Background and Significance

The Party Central Committee and the State Council attach great importance to the development of artificial intelligence. General Secretary Xi Jinping pointed out that artificial intelligence is a strategic technology that leads this round of technological revolution and industrial transformation, exhibiting a strong “leading goose” effect. In April of this year, the Central Political Bureau meeting emphasized the need to focus on the development of general artificial intelligence, create an innovative ecosystem, and pay attention to risk prevention. The recently concluded Central Economic Work Conference reiterated the need to vigorously promote new industrialization, develop the digital economy, and accelerate the development of artificial intelligence.
As a new stage in the development of artificial intelligence from specialization to generalization, the general artificial intelligence large model (hereinafter referred to as “large model”) integrates functions such as intelligent perception, intelligent analysis, intelligent decision-making, and intelligent execution. Through the deep integration of data, computing power, and algorithms, it achieves optimal allocation of production factors.
From the perspective of production structure, artificial intelligence technology has become a core component of modern industrial production, significantly improving production efficiency, optimizing resource allocation, and reducing production costs, providing strong technical support and impetus for new industrialization development. The deep integration of general artificial intelligence and manufacturing can further accelerate the industrial system’s advancement towards high-end, intelligent, and green development. To implement the decisions and deployments of the Party Central Committee and the State Council, and to promote the new generation of information technology to empower new industrialization, the China Industrial Internet Research Institute is preparing to establish an Innovation Center for the Integration of General Artificial Intelligence and Industry, focusing deeply on aspects such as large model foundations, architecture, standards, and applications, conducting a series of thematic research, summarizing and forecasting seven replicable and promotable new application models of large models in industrial scenarios, providing reference for the in-depth expansion of large models into manufacturing.

2. Overall Situation of Large Model and Its Industrial Application Development

On one hand, there is a global surge in the development of large models. The emergence of large models has pushed artificial intelligence into a new stage of development, with leading domestic and foreign enterprises becoming the technological high ground in the competition for large models. AI large models are artificial intelligence models with large-scale parameters and complex structures. With the increase in model size, the growing amount of training data, and the enhancement of computing power, AI large models have made significant breakthroughs in fields such as natural language processing, image recognition, speech recognition, and multimodal recognition. Since 2020, the market size of large models has rapidly grown globally, entering an explosive period. According to Precedence Research, it is estimated that by the end of 2023, the global large model market size will reach $13.7 billion, and by 2032, it will reach $118.1 billion. On the foreign side, OpenAI, as an industry-leading organization, has released language large models ChatGPT (2022) and GPT-4 (2023), speech large model Whisper (2022), and visual large model DALL-E (2021). Microsoft has integrated GPT-4-related capabilities into key products like Windows 11, Office 365, and Bing, forming the Copilot series applications; at the same time, Google launched the multimodal large model Gemini (2023), and Meta released the language large model LLaMA (2023), with foreign technology giants joining the large model competition one after another. Domestically, various technology companies are also actively following the global development trend of large models, with Baidu releasing the language large model “Wenxin Yiyan”, Alibaba releasing the language large model “Tongyi Qianwen”, iFlytek releasing the language large model “Xinghuo Cognition”, Baichuan Intelligent releasing “Baichuan Large Model”, Zhiyu AI releasing the ChatGLM series language large models, and the Chinese Academy of Sciences releasing the cross-modal large model “Zidong Taichu”.
On the other hand, the application of large models in the B-end, especially in the industrial field, has become an industry consensus. Large models have shown a development trend of using foundational large models as a technical base and industrial applications as an entry point, giving rise to the concept of industrial large models. The foundational large model integrates and expresses more domain knowledge and modal knowledge by enhancing the model’s parameter count and structural generality, forming an omniscient and omnipotent general model. The industrial large model, based on the structure and knowledge of the foundational large model, integrates data and expert experience from various industrial sectors, forming vertical, scenario-based, and specialized application models. Compared to foundational large models, industrial large models have advantages such as fewer parameters, higher specialization, and stronger feasibility, providing low-cost solutions for technological breakthroughs, product innovation, and production transformation in vertical industrial fields.
Current Status, Patterns, and Prospects of AI Large Model Industrial Applications

3. Seven Major Models of Large Model Industrial Applications

From the perspective of the entire lifecycle of industrial products, they can be divided into stages such as research and development, design, simulation, production, testing, operation and maintenance, and after-sales. This report analyzes the characteristics and elements of each stage and, in conjunction with the capabilities of large models, preliminarily proposes the following potential application models of large models in the industrial field.
Model 1: Principle-Based R&D
Large models can analyze product structural data to explore the configuration and mechanisms of products at a microscopic level, and generate products with new structures and new characteristics through the model’s emergent capabilities.
For example, in drug development, artificial intelligence large models can analyze large amounts of known drug molecule data to identify the best drug candidates and generate a new drug molecule design scheme, significantly shortening the time and cost of drug development and improving the success rate of drug development.
Target Discovery and Drug Structure Design. Target discovery is a core link in the drug development process, where the target is the binding site of the drug’s action in the body. The relationship between the drug and the target can be likened to a key and its corresponding lock. Traditional target discovery requires extensive scientific work, making numerous biological hypotheses about the target, and designing a series of experiments for verification, which takes a long time. However, large models can analyze known drug molecule structure data and knowledge graphs related to drug molecules and diseases to identify molecular features that interact with disease targets, and then automatically generate new drug molecule design schemes using these molecular features.
Drug Design Evaluation and Optimization. After generating several drug molecule design schemes, the model can evaluate the generated designs under human guidance, selecting the most promising drug candidates. It can also optimize drug efficacy and toxicity through the structure and properties of drug molecules, predicting drug metabolism pathways and drug concentrations to optimize drug dosages and administration schemes.
Assisting Clinical Trial Design. Experimental verification is an essential link in drug development, but experiments often consume a lot of time and resources. Artificial intelligence large models can help researchers design more effective clinical trial schemes, such as predicting the safety and efficacy of drugs, and optimizing the sample size and duration of clinical trials. By predicting and screening, the number and complexity of experiments can be reduced, thereby improving the efficiency and quality of drug development.
Model 2: Forward-Looking Design
Large models can generate innovative product design schemes, better assisting technicians in quickly translating design ideas and intentions into specific implementation plans.
For example, in traditional industrial design, large models can achieve rapid generation of engineering drawings and design proposals, and assist in layout optimization and parameter verification, significantly reducing the time spent on industrial design and enhancing product development efficiency.
Mathematical Modeling and Code Writing. In the industrial design process, many issues require the establishment of specific mathematical models for analysis and solution. For example, in mechanical product design, common problems include mechanical strength verification, size optimization, and standard parts selection, which can be costly to invite experts for analysis and modeling. Large models can provide reference schemes for these problems in a short time based on specific design needs, and automatically generate corresponding program codes to guide specific industrial design practices.
Diverse Design Scheme Generation. Taking CAD as an example, the existing massive standardized material libraries provide a wealth of data for engineering drawings and layout planning. Large models can leverage this data, combined with the designer’s creative ideas and special needs, to generate diverse design schemes for the designer’s reference. On the other hand, they can also rapidly optimize and adjust design schemes, helping engineers create layouts more quickly and with fewer errors.
Model 3: Efficient Simulation
Large models can utilize their generative capabilities to provide virtual simulation testing scenarios/environments that meet design needs, solving issues such as insufficient testing data and singular testing environments in industrial product design, thereby enhancing product reliability.
For example, in automobile manufacturing, artificial intelligence large models can generate simulation testing environments for conducting crash simulations and safety assessments, significantly improving the accuracy and efficiency of simulations and driving improvements in vehicle design and safety performance.
Diverse Simulation Scenario Generation. Automotive companies accumulate a wealth of real data during actual vehicle testing and simulation experiments, including vehicle structure, crash data, and material characteristics. Large models can learn the nonlinear relationships between vehicle structure, material properties, and crash responses using this data, understanding the impact of different parameters (such as crash speed, angle, and vehicle construction) on crash responses, and generating new, diverse crash scenarios through their emergent capabilities. This allows vehicle models to simulate predictions under various conditions, including previously unencountered situations, thereby compensating for data gaps in special scenarios and enhancing the comprehensiveness and accuracy of automotive simulation testing.
Parameter Optimization and Rapid Prediction. In traditional vehicle crash simulations, different parameter combinations require multiple rounds of testing to obtain the best results, with each test consuming significant time for manual parameter adjustment. However, large models can quickly predict crash response scenarios for different parameter combinations through zero-shot knowledge analysis capabilities in a question-and-answer format. This helps find the best parameter combination in a short time, reducing the cycle of vehicle design and testing. At the same time, based on historical crash cases, it can provide innovative design suggestions that may involve material selection, structural adjustments, and other aspects of innovation.
Model 4: Refined Detection
By leveraging the zero-shot learning capabilities of large models, combined with virtual reality technologies such as AR/VR, rapid and efficient visual detection of product quality defects, personnel violations, assembly errors, and other industrial scenarios can be achieved.
For instance, in quality inspection and safety monitoring in industrial production, through external visual sensing devices and simple command assistance, large models can detect specified areas and personnel based on demand, quickly identifying abnormal information and significantly reducing costs related to manual verification, sample collection, and model training.
High-Efficiency Industrial Quality Inspection. Taking PCB defect detection as an example, general visual large models can leverage strong generalization capabilities to perform semantic segmentation directly on raw PCB images without relying on factory sample data and localized fine-tuning training, identifying various defects such as solder voids, solder bridges, and pinholes by combining simple rules and methods. This addresses practical issues such as the difficulty of obtaining PCB sample data, the challenges of labeling, and avoids the high costs associated with training and parameter tuning, thereby improving the efficiency of defect detection in industrial products.
Intelligent Safety Production Supervision. In the coal mining industry, leveraging the machine vision recognition technology of large models, combined with equipment operation status data, can identify scenarios such as personnel entering hazardous areas, personnel falls, cutting tool placement, roof support, drilling depth, anchor agent usage, mixing time, secondary fastening, and anchor cable tension, transforming manual supervision of excavation operations into automatic monitoring, enhancing the standardization of excavation processes and improving the safety factor of coal mining production.
Personalized Detection Scenario Expansion. By combining language large models and visual large models, multi-modal perception and interaction can expand the application scope of large models in industrial visual inspection, enhancing the flexibility of applications. For example, in industrial quality inspection, voice commands can control large models to detect defects of different types, regions, and grades, meeting the inspection needs of various products.
Model 5: Intelligent Control
In large modern production lines, intelligent scheduling and control of multiple key nodes are necessary to enhance production line efficiency. Artificial intelligence large models can analyze diverse historical data to better understand the complex relationships in industrial scheduling tasks, such as production demand, resource availability, and task priority, thereby optimizing task distribution and scheduling at each node to improve production efficiency and flexibility.
Taking industrial robots as an example, large models can automatically integrate and analyze various production data, allowing for rapid task distribution and dynamic task adjustments for robots, becoming the “nervous center” of large industrial production lines.
Complex Pattern Learning and Rapid Task Distribution. In large-scale production lines such as automobile manufacturing, multiple industrial robots are usually configured to handle sub-tasks such as assembly, welding, and painting. Artificial intelligence large models can collect data on robot performance, workstation status, production plans, and based on this data, learn complex information such as robot skills, task complexity, and transfer times between workstations, predicting the efficiency of different robots executing different tasks. When a new task arrives, the model can quickly decide which robot to assign to minimize task waiting times and production cycles.
Dynamic Task Allocation Adjustment. Large models can analyze robot performance and efficiency from historical data and dynamically adjust task allocation strategies based on the requirements of different tasks. This helps optimize the use of robots, ensuring they perform optimally across different tasks. If issues arise such as robot malfunctions, workstation failures, or production plan changes, large models can quickly respond and reallocate tasks to address unforeseen circumstances.
Motion Control Code Generation. From the perspective of individual robot movements, production personnel can quickly generate customized motion control codes through text or voice interactions based on different task requirements, controlling robots to perform various tasks. For example, by inputting the command to the large model, “Please write a PLC program to control the robot to move parts from Point A to Point B.” This motion control instruction generation model based on large models can significantly enhance the flexibility of industrial robots, achieving flexible production line control.
Model 6: Scientific Operation and Maintenance
Large models can utilize their powerful reasoning capabilities to analyze and predict various data during the production process, thereby enhancing intelligent operation and maintenance levels and improving production management mechanisms.
For example, in warehouse management, large models can manage and integrate various categories and modalities of data in the supply chain, enhancing the operational efficiency of product information flow, creating a more convenient and faster industrial product supply chain system.
Intelligent Warehouse Management. Using robots embedded with multi-modal large models for shelf management, inventory management, and order picking can autonomously locate shelves, manage inventory, and transport items through the strong visual generalization capabilities of large models, enhancing warehouse operational efficiency. Additionally, using large models to predict inventory based on sales velocity and turnover rates can help formulate replenishment strategies, ensuring timely restocking to avoid sales impacts from inventory shortages while also preventing overstocking to avoid inventory buildup and capital occupation issues.
Efficient Data Management. The supply chain in industrial production involves a vast amount of data, such as raw material data, processing technology data, equipment status information, environmental information, personnel information, and logistics information. Large models can quickly organize, classify, and analyze data of different formats and sources, reducing data organization costs. Meanwhile, these classified and organized data can be further fine-tuned for large models, achieving a positive interaction between data and models.
Model 7: Customized After-Sales Service
Large models can leverage their significant advantages in natural language dialogue to make after-sales services no longer confined to fixed Q&A databases, allowing for more natural, fluent, and effective dialogues with customers, helping industrial enterprises achieve customized after-sales services that meet diverse user needs, thereby further enhancing customer loyalty and user growth, and expanding business scope.
For example, in after-sales service for machinery and equipment, large models can combine multi-modal and digital human technologies, allowing customers to describe equipment failures or issues in natural language, with the system accurately understanding and providing detailed, personalized solutions.
Multi-Dimensional Interactive After-Sales Service. The operation and maintenance of machinery and equipment require complex procedures, and traditional knowledge-based customer service systems cannot provide precise guidance to users. By utilizing large language models as backend logical reasoning support and virtual digital humans as frontend interactive figures, the system can accurately understand customer needs and quickly provide detailed, targeted solutions based on its knowledge reserves and specific issues. Additionally, through the gestures and voice interactions of virtual digital humans, customers can be assisted in operating equipment from multiple dimensions, improving the efficiency of after-sales services and providing customers with a more intuitive and personalized service experience.

4. Suggestions for Future Development

At this stage, constrained by fragmented industrial scenarios, insufficient computing resources, difficulties in collecting and organizing training data in the industrial field, and issues of safety and reliability of large models, the integration and application of large models in industry in China are still in the initial exploratory stage and face certain challenges.Firstly, foundational large models still dominate the application market and have yet to penetrate vertical industrial fields to form specialized industrial large models.Secondly, the current application of large models in industrial production is rather scattered, lacking standardized and systematic large model industrial application paradigms.Thirdly, building industrial pre-trained large models from the ground up has a high threshold, with only a few leading enterprises capable of conducting R&D on industrial large models. To address these issues, the following suggestions are proposed:
1. Promote Large Model Technological Breakthroughs for Industrial Scenarios. Identify common technological issues of large models aimed at industrial scenario applications, focusing on safety, reliability, and real-time performance, and encourage collaboration among universities, enterprises, and research institutions in specialized fields to tackle these issues. Strengthen domestic large model technological R&D through expanding industrial large model datasets, constructing typical industrial scenario rule sets, and optimizing model training algorithms to enhance the industrial application capabilities of large models.
2. Build a Scalable Industrial Large Model Data Resource Pool. Organize large model supply-side and demand-side enterprises and research institutions to develop training data specifications and testing standards for industrial large models. Relying on national industrial internet big data centers and other standardized platforms, establish an industrial large model corpus and form a management mechanism for industrial data. Through funding subsidies, tax reductions, and policy incentives, guide domestic medium and large manufacturing enterprises to share operational data from industrial production, forming a data resource pool covering key domestic industrial sectors to provide data support for the training and testing of industrial large models.
3. Improve the Performance Evaluation Mechanism for Large Models in the Industrial Field. Based on national authoritative institutions, collaborate with industry demand-side stakeholders to establish standardized large model industrial knowledge Q&A testing sets to ensure evaluation efficiency and result reliability. Additionally, establish a long-term performance evaluation mechanism for large models in the industrial field, periodically assessing key performance aspects such as knowledge capability, stability, and safety, dynamically adjusting evaluation indicators based on changes in industrial structure and data element distribution to promote the continuous empowerment of large models in new industrialization.
4. Promote Pilot Demonstrations of Typical Applications of Large Models in the Industrial Field. Integrate fragmented industrial scenarios, distilling typical business scenarios of large model industrial applications represented by product-assisted design, refined quality inspection, and intelligent supply chain management, clarifying quantitative demand indicators for large models in each scenario, and promoting the establishment of relevant industry standards. Establish a two-way interaction mechanism between large model supply-side and enterprise application-side to promote the formation of several characteristic industrial clusters for the collaborative development of large model R&D and manufacturing, and encourage the establishment of benchmark and exemplary applications of industrial large models.

END

Current Status, Patterns, and Prospects of AI Large Model Industrial Applications
Current Status, Patterns, and Prospects of AI Large Model Industrial Applications

Leave a Comment