AI Expansion into the Manufacturing Industry: Use Cases and Future Prospects of AI in Manufacturing

1. Background of AI Utilization in Manufacturing

It is no exaggeration to say that the manufacturing industry continuously pursues improvements in production efficiency and quality. The bottom-up activities unique to Japanese companies, known as "Kaizen," have been adopted overseas as a model and have long driven the competitive advantage of Japanese manufacturing. However, it is also true that due to the shortage of skilled workers, soaring raw material prices, the complexity of supply chains, and intensified global competition, the limits of traditional improvement methods are beginning to show. In this context, the utilization of AI (artificial intelligence) has attracted attention. For example, identification-type AI has the ability to extract patterns from vast amounts of data on the manufacturing floor and make judgments and proposals, and its introduction has begun in various stages of the manufacturing process. This time, we will divide AI into "identification-type AI" and "generative AI" and introduce specific examples of their utilization.

▼Related Blog
Use Cases and Benefits of Image Judgment AI for Manufacturing Industry

2. Differences Between Discriminative AI and Generative AI

Until now, AI generally referred to discriminative AI, but with the advent of generative AI, the term discriminative AI has come to be used to distinguish it from the traditional AI. First, let's clarify the fundamental differences between the two.
To put it simply, discriminative AI acts as a substitute or support for the "eyes and ears on the ground," while generative AI substitutes or supports "knowledge and document creation on the ground." Their main uses and areas of expertise differ, with discriminative AI already being widely utilized in many manufacturing sites. In contrast, generative AI is often seen being used by staff supporting the site and in indirect departments.

●Comparison between Discriminative AI and Generative AI

　 　 　 　 　 　 　 　 　 　 　 　 　 　 　

	Discriminative AI	Generative AI
Main Role	To "discriminate and classify" data	To "generate" data
Main Purpose	Quality inspection, machine and robot control, demand forecasting (production planning optimization), anomaly detection, etc.	Automatic document and manual creation, design support, inquiry and dialogue support, etc.
Input Data	Sensor values, images, audio data, etc.	Text, specifications, design data, etc.
Implementation Effects	Improvement of inspection accuracy and reduction of man-hours, risk of production stoppage, inventory reduction	Acceleration of development speed, efficiency improvement of on-site support operations (such as reduction of document preparation man-hours), support center efficiency

3. Use Cases of Discriminative AI in Manufacturing

Identification AI uses object recognition, region detection, and pattern recognition technologies, and in manufacturing sites, it analyzes cameras, sensors, and various data to be utilized for "discrimination," "classification," and "prediction." Specifically, there are the following use cases.

●Automation of Quality Inspection
By automating visual inspections with AI, it has become possible to accurately detect abnormalities that previously required skilled inspectors. Many visual inspections rely on human eyesight, and challenges include inspection accuracy depending on the inspector’s proficiency and frequent variability. By introducing AI image recognition, there are increasing cases where even minute scratches and foreign object contamination can be detected with high precision.
Toyota Motor Corporation has introduced WiseImaging, an AI from CEC, automating magnetic particle inspection that cannot be detected visually, significantly improving inspection accuracy while reducing the labor-intensive and time-consuming inspections that previously required skilled personnel.

●Automation of Skilled Manufacturing Machine Control
AI that analyzes data from sensors and performs machine control requiring skill and expertise is becoming more widespread.
Yokogawa Electric has introduced AI-based autonomous control in chemical plants, enabling AI to control areas that previously could only be managed manually. This has successfully eliminated losses in fuel, labor costs, and time caused by the occurrence of nonconforming products.

Bridgestone has developed and released "EXAMATION," a tire molding system equipped with AI, and has already succeeded in automating processes from material handling to inspection. By minimizing various human-induced variations, it not only enables unprecedented high-precision manufacturing but also contributes to high productivity and skill-less operation through automation.

●Automation of Demand Forecasting
There are increasing cases of using AI to advance demand forecasting and overall factory optimization.
Sapporo Breweries conducted a verification of demand forecasting through collaboration between humans and AI. As a result, the forecasting accuracy improved by 20% compared to human-only forecasts, and operations have begun. It is expected to greatly contribute to planning and execution across the entire supply chain as well as inventory optimization.

4. Use Cases of Generative AI in Manufacturing

It goes without saying at this point that generative AI is a technology that generates text, images, and various types of data. In the manufacturing industry as well, the utilization of generative AI is opening up new possibilities, with adoption and consideration progressing. It is expected to contribute not only by generating text but also by producing work manuals, work instructions, design data, and more, thereby improving operational efficiency and accelerating development speed.

●Design and Development
At the product development stage, generative AI automatically generates new design proposals based on existing data and specifications, expanding engineers' creativity.
Panasonic HD utilizes generative AI to optimize the design of motors for electric shavers. Until now, improvements were made through the experience of skilled engineers, but they have developed a zero-based design method using advanced algorithms powered by generative AI. Compared to motors optimally designed by experienced technicians, this approach achieves a 15% increase in output performance, balancing performance improvement with shorter development periods, and is expected to become mainstream in future equipment design.

●Inquiry Response (Customer Support)
Generative AI is also being utilized in customer support.
Fujitsu has introduced Salesforce's "Einstein for Service", advancing the use of automatic reply generation for inquiries and summarization of conversations between customers and operators. Functional verification reported that automatic reply generation reduced customer response time by approximately 89%, and conversation summarization cut the time required for post-interaction record-keeping by about 86%.

5. Future Challenges and Prospects

As has been stated so far, there is no doubt that AI holds diverse possibilities. However, while the media and the internet tend to emphasize only the positive aspects, risks do exist in both discriminative AI and generative AI.

In discriminative AI, misjudgments can occur due to biases in training data, leading to errors in handling unexpected patterns, or mistakenly classifying anomalies and noise as normal. Since high-precision AI requires large volumes of high-quality data, it is necessary to establish data infrastructure and implement verification processes for data collection and cleansing. Moreover, overreliance on AI can result in critical defects being overlooked, so ensuring data diversity and establishing a final human check system are indispensable.

In generative AI, there are risks such as hallucinations that present information differing from facts, information leakage, and copyright infringement. Therefore, human checks, on-premises local LLMs, and the consideration of guidelines for the use of generative AI are also necessary here.

Going forward, the key to enhancing corporate competitiveness through the use of generative AI lies not only in the "skillful utilization of AI" but also in the "quality of risk governance." Additionally, it is necessary to realize new manufacturing sites through the collaboration of humans and AI by appropriately leveraging both discriminative AI and generative AI. It is not difficult to imagine that fostering human resources who understand both AI technology and the site-specific technologies and operations will be crucial for this.

6. Human Science Teacher Data Creation, LLM RAG Data Structuring Agency Service

Extensive Track Record of Creating 48 Million Pieces of Training Data
At Human Science, we participate in AI model development projects across a wide range of industries, starting with natural language processing and extending to medical support, automotive, IT, manufacturing, and construction. Through direct business with many companies, including GAFAM, we have provided over 48 million pieces of high-quality training data. From small-scale projects with just a few members to large long-term projects with a team of 150 annotators, we handle various types of annotation, data labeling, and data structuring regardless of industry.

Resource Management Without Using Crowdsourcing
At Human Science, we do not utilize crowdsourcing; instead, we advance projects with personnel directly contracted by our company. We carefully assess each member's practical experience and their evaluations from previous projects to form a team that can deliver maximum performance.

Generative AI LLM Dataset Creation and Structuring, Also Supporting "Manual Creation and Maintenance Optimized for AI"
We handle not only labeling for data organization and creation of training data for discriminative AI, but also the structuring of document data for generative AI and LLM RAG construction. Since our founding, manual production has been a core business and service, and currently, we also support "the organization of business knowledge and manualization for future generative AI and RAG introduction and utilization." We provide optimal solutions leveraging our unique expertise in the structures of various documents.

Equipped with an In-House Security Room
At Human Science, we have a security room within our Shinjuku office that meets ISMS standards. Therefore, we can guarantee security even for projects handling highly confidential data. We consider ensuring confidentiality to be extremely important in every project. Even for remote projects, our information security management system has received high praise from clients, as we not only implement hardware measures but also continuously provide security training to our personnel.

In-House Support
Our company also provides personnel dispatch services for annotation-experienced staff and project managers who match the customer's tasks and situation. It is also possible to organize a team under the customer's on-site supervision. Additionally, we support the training of your workers and project managers, selection of tools tailored to your situation, automation, work methods, and the construction of optimal processes to improve quality and productivity, assisting with any challenges you face related to annotation and data labeling.