Context is everything – new project develops industry-friendly AI
Photo: Unsplash.
Many industrial companies find it difficult to use general-purpose AI language models in their operations. In a new research project at Jönköping University, researchers will therefore develop AI solutions that combine language models with industry-specific knowledge to make the technology more accurate, comprehensible and usable in real-world manufacturing environments.
“I hope the results will contribute both to companies’ digital transformation and to scientific development within knowledge-intensive and explainable AI,” says He Tan, Senior Lecturer in Computer Science at the School of Engineering at Jönköping University and the project leader.
By giving AI access to clear and structured information about how the industry works, the systems can better understand context. This enables AI to provide more accurate answers, become easier to understand, and demonstrate in a clearer way why it is giving a particular piece of advice.
The project is being carried out in collaboration with industrial partners, including the industrial company Comptech and the IT consultancy firm Consid.
“We see this project as an important step towards more intelligent and knowledge-based manufacturing,” says Per Jansson, CEO of Comptech.
The project aims to make AI more useful in industrial environments by increasing understanding of the specific context in which the work is carried out.
“Today, many AI solutions are powerful, but they often lack a deep understanding of the industrial context in which they are to operate. With this new project, we want to bridge that gap,” says He Tan.
Explainable AI refers to AI systems that not only provide an answer, but can also explain how and why they arrived at that answer.
From raw data to usable knowledge
The manufacturing industry generates vast amounts of data in the form of logs, reports, sensor data and inspection results. At the same time, this information is often disjointed and difficult to use directly in advanced AI systems. In this project, researchers will develop methods to transform raw data into structured, machine-readable knowledge that can then be used in conjunction with language models.
The project has four overarching objectives: to extract industrial knowledge from various types of data, to structure this knowledge into models, to integrate it with language models, and finally to test the solutions in practical industrial applications using AI agents.
“By combining AI language models with structured industrial knowledge, we can create AI systems that not only provide answers, but can also explain how they reason,” says He Tan.
The new project addresses several of the biggest challenges facing today’s industrial AI, such as a lack of transparency, limited domain knowledge and difficulties in using the models in real.
