However, the uncertainty of demand and supply makes it difficult for traditional management methods to help companies achieve efficient production, while AI has the ability to consider a large number of variables at the same time and can learn to schedule and act as an agent to enable companies to maximize profits.
The complexity of current industrial products and operating systems is increasing dramatically, making the traditional model of relying on experienced engineers to spend a lot of time on experimentation and research extremely inefficient. At the same time, the reputation of industrial enterprises is based on product quality, and innovation is a key factor for the sustainable growth of enterprises. Therefore, industrial enterprises need to quickly understand and solve problems, and AI can break the traditional problem-solving paradigm of enterprises, help enterprises accelerate business operations, and create huge value for enterprises.
For example, to find problems in a system, AI can use causal modeling to reformulate complex problems into specific business problems, build event-based data models, and link thousands of variables in the product development lifecycle and operational history, including design configurations, manufacturing parameters, and maintenance and repair history. Then, based on the model, it can identify the highest factors related to a specific problem, allowing engineers to quickly find the most likely root cause of the problem.
At the end of 2022, ChatGPT, a brand-new AI chatbot released by the AI research lab OpenAI, quickly gained popularity due to its groundbreaking "Transformer architecture large model + RLHF (Reinforcement Learning from Human Feedback) algorithm," which brought about a huge improvement in the human-computer interaction experience in terms of logical expression and naturalness of expression. It also brought more possibilities for improving the efficiency and quality of products throughout their entire life cycle in the industrial field.
Applications of ChatGPT in the industrial sector
2.1 Development History of General AI and Industrial AI Technologies
The differences between industrial AI and general AI like ChatGPT lie in two aspects. First, high-value applications of industrial AI are typically concentrated in scenarios strongly integrated with industrial mechanisms, such as predictive maintenance of equipment, production process control optimization, knowledge-based comprehensive decision-making, and visual inspection of defective products. The distribution of industrial AI application scenarios is shown in Figure 1. Second, the high precision, professionalism, and security requirements of AI in industry lead to a time lag between technological innovation and practical industrial application. For example, the industrial application lag cycle for statistical machine learning is generally around 10 years. New technologies such as deep learning and generative adversarial networks were applied in general fields after 2012, and industrial applications emerged four years later. Overall, due to the increased availability of AI technology and the improvement in automation, informatization, and intelligence levels in the industrial sector, the breakthrough of the general AI model brought by ChatGPT is still immature for industrial applications in the short term, but it is expected to experience rapid development in the medium to long term.
2.2 Short-term application limitations in the industrial sector
In the short term, ChatGPT's ambiguity, lack of timeliness, non-open source nature, lack of professionalism, low credibility, and questionable intellectual property rights limits its application and iteration in the industrial field.
1) High degree of ambiguity. The accuracy and precision of ChatGPT's information retrieval and collection need to be improved, which is key to ChatGPT's commercialization in the industrial field.
2) Low timeliness. The timeliness of ChatGPT is affected by the update frequency of OpenAI's model database and the data source, which is insufficient to support the requirements of industrial enterprises for the immediacy and predictability of information, and there is a risk of insufficient information timeliness.
3) Non-open source nature. ChatGPT is not an open source system, and the current training data is public domain data. Its application in closed enterprise environments is unknown, and most industrial enterprises do not yet have enough high-quality, accurate (correct) basic data for model training.
4) Lack of professionalism. ChatGPT is a general language model. Its application in the industrial field still requires appropriate methods to combine it with industrial knowledge. That is, even a student who has completed nine years of compulsory education still needs some professional skills and qualities to carry out work practice.
5) Low credibility. The current quality of content generated by ChatGPT does not yet meet the high credibility requirements of information and data in the industrial sector.
6) Low security. ChatGPT has data leakage and privacy protection issues during its application within industrial enterprises. Due to concerns about information leakage and other problems, companies such as SoftBank and Hitachi have begun to restrict the use of interactive AI services such as ChatGPT in their commercial operations.
7) Intellectual property issues are questionable. The intellectual property issues of ChatGPT need to be discussed. The ownership and copyright protection of new designs or ideas proposed by generative AI are questionable.
2.3 Potential medium- to long-term applications in the industrial sector
In the medium to long term, ChatGPT has considerable application potential in areas such as disrupting industrial software, solving industrial field problems, and building knowledge graphs.
First, it will revolutionize the form of industrial software. The essence of software lies in data collection and aggregation, information use, and the control of human operational processes. With the increasing diversification of production methods, ChatGPT is likely to revolutionize the future form of software, potentially transforming from an app or desktop application into an interactive form. The information flow will become "human-machine-IoT device-ChatGPT." ChatGPT will meet the requirements of different roles regarding the granularity, type, and real-time nature of information generation and delivery in a more intelligent way. It will not only explore diverse innovations or design solutions in a short time to improve efficiency, but also rapidly iterate and optimize based on material, manufacturing process, and performance parameter requirements. The generative design capabilities and process simulation levels of industrial software, represented by CAD and CAE, are expected to be further enhanced. AI-enabled simulation optimization and digital twin optimization are shown in Figure 2.
Secondly, it addresses problems in industrial settings. After decades of operation, companies often possess a wealth of data. Therefore, when engineers troubleshoot highly complex systems, they need to find the most relevant historical programs, machine performance data, and operational examples. However, it's nearly impossible for humans to browse millions of structured and unstructured data records to obtain information. In this situation, ChatGPT promises to help engineers accelerate problem-solving by ingesting large amounts of data and quickly locating the most relevant information.
For example: ① Clarifying requirements and writing code. When faced with programming needs in industrial settings, engineers can use ChatGPT as a "translator," writing the necessary code to solve problems in the industrial environment once the requirements are clear. ② Information retrieval and analysis. ChatGPT acts as a "guide," importing resources such as the standards library of the National Standardization Administration, the literature database of the Machinery Industry Information Research Institute, and the company's own knowledge base into its basic model training database. This allows for information retrieval, navigation, and analysis, improving information retrieval efficiency. ③ Enhancing industrial manufacturing efficiency. ChatGPT can construct bills of materials (BOMs) from data obtained from product configuration, development, and procurement, identifying opportunities for reusing parts and improving existing work standards, helping companies shorten engineering production time. Using ChatGPT to improve the computational efficiency of MRP (Material Requirement Planning) and APS (Advanced Planning & Scheduling System) can result in at least a 5% performance improvement in industrial manufacturing. ④ Improving the ease of software development and use. On the one hand, ChatGPT acts as a "shortcut key," quickly accessing specialized tools in systems such as SAP and MES, enhancing the convenience for system software users. On the other hand, ChatGPT acts as an "assistant," helping industrial software users model and develop applications faster, more consistently, and with higher quality. For example, using Siemens' low-code application development platform Mendix, users can leverage MxAssist Logic Bot (an AI-driven virtual collaborative developer robot) to build the required application logic, complete modeling and configuration faster and with higher quality on the platform, without having to write code.
Thirdly, ChatGPT provides online answers to its application scenarios in industry, including cross-industry popularization, engineering and technical support, knowledge updates, data analysis and language translation, and online customer service. ① Cross-industry popularization: ChatGPT can answer various science and technology, industrial, and other related questions, quickly simplifying complex technical knowledge into an easy-to-understand format, thus helping people outside the field understand the area. ② Engineering and technical support: ChatGPT can help engineers find code examples and technical solutions more quickly, accelerating the development process. For example, it can generate code based on the requirements provided by engineers, provide code modification suggestions, and highlight best practices and common errors during the writing process. ③ Knowledge updates: ChatGPT can help employees of industrial companies quickly learn and update their technical knowledge. For example, it can provide the latest technical information and research progress based on the requirements provided by engineers. ④ Data analysis: ChatGPT can utilize large amounts of data for analysis, helping industrial companies better identify patterns and trends and make decisions based on this. In addition, ChatGPT has many other application scenarios, such as helping industrial companies with language translation, document generation, speech synthesis, online customer service, and providing prior knowledge/experience references.
The impact of ChatGPT on employment
From the perspective of the overall impact of AI technology applications, AI technology brings both opportunities and challenges in creating and maintaining high-end jobs: on the one hand, research shows that AI technology may have a positive impact on wage growth for workers with highly specialized skills or higher levels of education; on the other hand, the popularization of AI technology may lead to the replacement of some jobs by gig work, resulting in a decrease in demand for permanent jobs and an increase in non-standardized freelance employment opportunities.
From the perspective of ChatGPT's impact on industrial jobs, in the short term, ChatGPT is merely a generalist with a score of 60 in all subjects, lacking specialized industrial knowledge and practice, and therefore will not lead to large-scale job replacement. In the long term, ChatGPT will inevitably bring about efficiency improvements, leading to adjustments in resource structures, including adjustments and iterations in job structures. For example, while robotic arms have replaced some assembly workers, they have also created positions related to automation and digital operations; since the launch of AI products, a series of positions such as training, application, development, and maintenance have emerged; and after the advent of automobiles, according to statistics, the number of people engaged in transportation services has increased thousands of times.
Therefore, in order to address the application of ChatGPT, industrial enterprises need to create new, highly efficient jobs. In the long term, ChatGPT is expected to replace some tasks in human work, but it cannot completely replace all jobs. The impact of ChatGPT will depend on the investment and technology ownership models of policymakers and research institutions.
Conclusion
ChatGPT, as an AI technology, has democratized content generation. However, due to limitations such as the ambiguity of its information, the low reliability of its data, and the immaturity of its business model, its application in the industrial sector is not yet mature in the short term, and its impact on industrial enterprises is limited. In the long term, however, ChatGPT shows promising prospects in the industrial field, acting as a "translator," "guide," and "shortcut key" in solving industrial problems, and possessing value in improving industrial efficiency.
Meanwhile, almost every technological revolution is accompanied by adjustments in the structure of labor factors. During Industry 3.0 and 4.0, "machines replacing humans" has been a recurring hot topic. The recent viral success of ChatGPT has further drawn public attention to the structural adjustments in certain job positions. Therefore, before content-generating technologies truly impact industrial jobs, companies should proactively plan for, add to, and train personnel for highly specialized positions to prepare for the inevitable future impacts on employment and business operations resulting from structural adjustments.
