Development of Generative AI and Its Applications in the Manufacturing and Machinery Industry

Generative AI Has Broad Industrial Application Potential and Benefits

Overview of Generative Artificial Intelligence
Generative Artificial Intelligence (Generative AI, or GenAI) is a type of AI technology based on machine learning and deep learning. After training models with large volumes of data, GenAI can autonomously generate new content based on user input.

For example, the highly popular GenAI model ChatGPT uses the Generative Pre-trained Transformer (GPT) approach, leveraging massive amounts of unlabeled, unstructured data and self-supervised learning to build a foundation model. These foundation models can automatically generate new content based on input data.

However, for specialized domain problems (such as specific engineering or industrial applications), foundation models may not produce satisfactory answers. In such cases, a small amount of labeled data can be used to fine-tune the foundation model into a domain-specific model for applications in finance, design, robotics control, and more.

Some multimodal GenAI models support input and output in various formats, including text, speech, images, video, code, and 3D model data. Input and output types can differ—for example, using text prompts to generate images or videos.

Capabilities and General Applications of Generative AI

Current GenAI models offer the following capabilities and applications:

• Intelligent Dialogue: Supports natural language conversations and basic Q&A, useful for human-computer interaction scenarios such as smart customer service.
• Document Creation and Translation: Generates text content from prompts, writes essays or letters, summarizes documents into key points or tables, performs automatic and real-time translation, and provides multilingual dialogue with improved accuracy.
• Code Generation: Automatically generates code from natural language or image descriptions, or repairs and adjusts existing code.
• Decision Support: Quickly consolidates key information from large datasets, analyzes data to propose solutions, and evaluates or optimizes strategies.
• Knowledge Management: Integrates large volumes of unstructured data, enabling users to retrieve information or perform knowledge searches via natural language dialogue.
• Design Generation: Creates images, videos, or even 3D models based on text or sketches, and performs image synthesis, editing, and special effects rendering.

Challenges in Applying Generative AI

Despite its potential, GenAI faces practical challenges:

• Model outputs are strongly tied to training data. Limited and imperfect data can lead to logical or factual errors. Biases in training data may increase risks in generated content.
• Results cannot guarantee full accuracy or usability, especially in sensitive domains like safety, law, health, ethics, finance, and politics. Additional human or automated review and fine-tuning with domain-specific data are necessary.
• Building foundation and specialized models requires massive data, time, and computational resources, incurring high costs—commercial foundation models average around $100 million to develop.
• Intellectual property issues arise regarding training data, user data protection, and ownership of generated content, requiring technical and regulatory solutions.

Applications in Manufacturing and Machinery

GenAI offers significant potential in manufacturing and machinery, including:

• Administrative Tasks: Business correspondence, marketing copy, contract drafting and review, document translation, meeting minutes generation—tasks that general-purpose GenAI models can handle to boost efficiency.
• Specialized Tasks: Require domain-specific models and data, such as:
  • Product Design Optimization: Generate prototypes from text prompts, refine designs iteratively, and integrate with CAE tools for engineering compliance.
  • Robotics Control: Enable natural language commands for robots, auto-generate control code, and use multimodal models for environment-aware actions.
    • Examples: Google and TU Berlin’s PaLM-E model for vision-language integration; Japan’s Denso using ChatGPT to generate industrial robot control code.

Benefits for Manufacturing Efficiency

According to BCG, GenAI can:

• Improve operational transparency and assist workers.
• Support decision-making for stability and predictability.
• Enable autonomous systems for structural optimization and adaptability.

Taiwan’s GenAI Development and Industry Applications

• National Science Council: Developing TAIDE, a trusted AI dialogue engine tailored to local culture and language.
• Taiwan AI Cloud (TaiZhiyun): Offers Formosa Foundation Model (FFM) and AI Foundry Service for enterprise deployment.
• MediaTek: Launched DaVinci platform for customizable GenAI applications.
• Tool Machine Industry:
  • NTU team: AR-based service platform, inspection assistance, automated quoting systems.
  • Local companies: Embedded AI modules (Mr.Q), GenAI chatbots (T-BOT) for machine operation and troubleshooting.
• Government Initiatives: Ministry of Economic Affairs promoting GenAI adoption in textiles and machine tools through pilot projects.

Enterprise Strategies for GenAI Adoption

Companies should:

• Assess GenAI’s value (efficiency, cost reduction, innovation) and risks (cost, security, IP).
• Establish risk management frameworks and ethical oversight.
• Choose among three implementation models:
  1. Build proprietary foundation models (high cost, full control).
  2. Fine-tune existing models (moderate cost, faster deployment).
  3. Use general models with prompt engineering (lowest cost, higher security risk).

Conclusion

GenAI has immense potential to transform manufacturing and machinery industries by driving innovation, efficiency, and competitiveness. However, its adoption requires systematic evaluation, robust ecosystems, and cross-industry collaboration to address technical, security, and regulatory challenges.