Targeted AI Enhancements Promise Gains in Product Quality and Worker Safety

According to a new study co-authored by a cross-disciplinary team of experts from the University of Notre Dame, targeted AI enhancements in manufacturing and the service industry can lead to improved product quality and enhanced worker safety. The study was published in Information Fusion

artificial intelligence

Image Credit: everything possible/Shutterstock.com

Recent progress in artificial intelligence has predominantly centered on text-based applications. AI increasingly demonstrates potential in diverse areas, including manufacturing and the service sector.

The study investigates the potential impact of a category of AI tools capable of processing various types of input and performing reasoning on the future of work. These tools, which include models like ChatGPT, are known as multimodal large language models.

While most research on AI and employment has concentrated on office-based tasks, this new study examined production work environments, where the advantages of AI might appear less obvious.  

Notre Dame researchers collaborated with welding professionals in Indiana from the Elkhart Area Career Center, Plymouth High School, Career Academy South Bend, Plumbers & Pipefitters Local Union 172, and Ivy Tech Community College to collect images for their study. These collaborations were facilitated through the relationships established by the University’s iNDustry Labs. Northern Indiana has one of the highest concentrations of manufacturing jobs in the United States, and iNDustry Labs has partnered with over 80 companies in the region on more than 200 projects.  

The research focused on welding across several industries: recreational vehicle and marine, aeronautical, and agricultural. The study assessed the accuracy of large language models in evaluating weld images to determine their suitability for different products. The researchers discovered that while these AI tools showed potential in assessing weld quality, their performance was significantly better when analyzing carefully selected online images compared to real-world weld examples.

This discrepancy underscores the need to incorporate real-world welding data when training these AI models, and to use more advanced knowledge distillation strategies when interacting with AI. That will help AI systems ensure that welds work as they should. Ultimately, this will help improve worker safety, product quality and economic opportunity.

Nitesh Chawla, Study Co-Author, the Frank M. Freimann Professor, Computer Science and Engineering, University of Notre Dame

Chawla is also the Founding Director of the University’s Lucy Family Institute for Data and Society.

The researchers observed that providing prompts tailored to specific contexts could, in some instances, improve the performance of AI models. They also noted that the size or complexity of the models did not consistently correlate with enhanced performance. Ultimately, the co-authors of the study suggested that future research should prioritize improving the ability of these models to reason effectively in unfamiliar or novel situations.

Our study shows the need to fine-tune AI to be more effective in manufacturing and to provide more robust reasoning and responses in industrial applications.

Grigorii Khvatski, Doctoral Student and Lucy Family Institute Scholar, Department of Computer Science and Engineering, University of Notre Dame

Yong Suk Lee, an Associate Professor of Technology, Economy, and Global Affairs at Notre Dame’s Keough School of Global Affairs and the program chair for technology ethics at Notre Dame's Institute for Ethics and the Common Good, stated that the study's findings have significant implications for the future of employment.

As AI adoption in industrial contexts grows, practitioners will need to balance the trade-offs between using complex, expensive general-purpose models and opting for fine-tuned models that better meet industry needs. Integrating explainable AI into these decision-making frameworks will be critical to ensuring that AI systems are not only effective but also transparent and accountable.

Yong Suk Lee, Associate Professor, Technology, Economy, and Global Affairs, University of Notre Dame

The study was supported by funding from the U.S. National Science Foundation Future of Work program and is among the federally funded research initiatives at the University of Notre Dame.

In addition to Chawla, Khvatski, and Lee, the study's co-authors include Corey Angst, the Jack and Joan McGraw Family Collegiate Professor of IT, Analytics, and Operations in the University’s Mendoza College of Business; Maria Gibbs, the senior director of Notre Dame’s iNDustry Labs; and Robert Landers, an Advanced Manufacturing Collegiate Professor in Notre Dame’s College of Engineering.

Journal Reference:

Khvatskii, G., et al. (2025) Do multimodal large language models understand welding? Information Fusion. doi.org/10.1016/j.inffus.2025.103121

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.