Page 50 - IJAMD-1-2

International Journal of AI for
Materials and Design
A unified ILKM in smart manufacturing

3.4. Intelligent domain expert ML system specific industrial tasks, utilizing relevant structured

Upon the successful training of the domain knowledge industrial data and domain-specific knowledge to provide
LLM, the fourth step involves utilizing it as a domain precise, expert-level solutions. In contrast, LLMs are more
expert for subsequent specialized model development. generalized, leveraging extensive training on diverse
textual data to solve language-related tasks, such as text
In this step, domain instruction data serves as the generation, conversation, and language translation. To
prompt, propelling the LLM to address specific analytical better illustrate the characteristics of ILKMs, a detailed
problems. The domain-specific knowledge LLM, acting comparison between ILKM and LLM is presented in
on these instruction inputs, proposes targeted solutions. Table 1. They are compared and explained from eight
In addition, human experts may interact and intervene, perspectives: “Data,” “Purpose,” “Applications,” “Data
offering strategic guidance to refine the LLM’s outputs. Privacy and Security,” “Domain-Specific Knowledge,”
These solutions are then transferred to a coding-focused “Integration and Customization,” “Scalability,” and “Real-
LLM, 25,26 which incrementally develops code aligned with Time Decision-Making.”
the domain knowledge LLM’s insights, thereby creating
a new ML model for specific problems. In addition, the 4.2. Foundational principles for ILKM development
structured machine-generated data serves as the dataset As illustrated in Figure 3, the “6S Principle” is proposed as
for new ML model training and testing. Finally, this step a guideline for the future development of ILKMs. The “6S
culminates in the generation of actionable solutions, ready Principle” encompasses six key components: “Specialized
to be integrated into decision-making workflows. Domain Knowledge,” “Scrutability,” “Safety,” “Scalability,”
4. Discussion “Sustainability,” and “Systematization and Standardization.”
The details of the purpose, challenges, and opportunities
This section discusses the comparison between ILKMs for each principle are presented in Figure 3. All of these
and LLMs and introduces the “6S Principle” as a guideline principles are crucial for the successful application of
for future ILKM development. It also highlights several ILKMs in industrial settings, ensuring that ILKMs can
potential opportunities for ILKM deployment in Industry address specific needs and challenges faced in Industry 4.0
4.0 and smart manufacturing. and smart manufacturing.

4.1. Comparison between ILKMs and LLMs 4.3. Prospective and opportunity
The main difference between ILKMs and LLMs lies in their There are several opportunities for developing ILKMs in
purpose and functionality. ILKMs are designed to handle the future of Industry 4.0 and smart manufacturing. In the

Table 1. Comparison between ILKMs and LLMs

ILKM LLM
Data Industrial domain-specific data (human-interpretable data and Vast, diverse, and unstructured text data; public open source
structured machine-generated data); private closed source
Purpose Designed for specific industrial tasks; provide specialized Designed for language-related tasks; focus on understanding
solutions in respective domains and generating human language
Domain-specific Specialized: in-depth, domain knowledge relevant to specific General: may lack deep, industry-specific insights
knowledge industries
Data privacy and Offer greater control over data privacy and security as they can Potential concerns with data privacy and security as
security be hosted within the company’s secure environment researchers often use licensed pre-trained models developed by
other private companies to fine-tune LLMs
Integration and Tailored and integrated into a growing and evolving industrial Need additional resources for integration and customization to
customization ecosystems, aligning with industry-specific needs fit specific requirements
Scalability Adapt and expand based on specific industrial requirements and Scalable across platforms but also requires significant
environment, but need to be balanced with the cost computational resources
Real-time Better suited for real-time decision-making in industrial settings, Limited in handling real-time, complex industrial decisions
decision-making leveraging specific industry data due to generic training
Application Process optimization, predictive maintenance, quality control, Text generation, content creation, conversation, language
prognostic and health management, material and design, data translation, summarization, etc., Not domain specific
analytics, decision-making, question-answering platforms, etc.
Abbreviations: ILKM: Industrial large knowledge model; LLM: Large language model.

Volume 1 Issue 2 (2024) 44 doi: 10.36922/ijamd.3681