There are still barriers blocking AI success in manufacturing which means that insights from connected systems are not delivering the expected benefits in performance and quality. As Professor Rab Scott from High Value Manufacturing Catapult, and Nick Wright, Director of Market Development at Digital Catapult explain, the fault does not lie with technology.
It is a big question and a concern that often arises when engineers discuss the use of AI in the design and manufacturing of products. While the underlying AI technology is fundamentally important, success hinges on addressing systemic issues that obstruct the successful application in industry.
Digital Catapult and the High Value Manufacturing Catapult are collaborating with industry, academia and government to drive the adoption of deep tech innovation to solve complex challenges in the manufacturing sector. Together, they have identified why the benefits in performance and quality aren’t delivering, and the importance of innovation support for the sector.
Data readiness
Many organisations still face challenges with data being stored in disparate systems or silos, where access (from both a technical and governance perspective) and availability are fundamental challenges. This fragmentation hinders the ability of AI to analyse complex comprehensive multimodal data sets – essential for generating connected insights.
In manufacturing, data is generated and used across various functions including production, quality control, inventory management and the supply chain. Innovation in AI that is focused on data access and integration is required to enable higher value outcomes from more developed AI applications. Another challenge is that while AI models require high-quality, accurate and relevant data to generate insights of value, they are often given poor data quality, incomplete datasets and outdated information which can lead to incorrect predictions or suboptimal recommendations.
Propriety (and availability) of vital training data
The advancements in generative AI (e.g., ChatGPT) are largely due to the extensive scale and diversity of the datasets used for training, combined with significant computational resources and innovative model architectures. These have enabled impressive fluency in general discourse and the development of techniques such as retrieval-augmented generation, which allow models to adapt to specific use cases. However, in complex domains requiring deep, specialised knowledge such as manufacturing, AI outputs can often be misleading or confused.
This can be overcome through the use of domain-specific datasets and targeted model training. Unlike the web-scraped data typically used to train large language models (LLMs), the specialised data required in these fields is often closely guarded as proprietary trade secrets, making it unlikely to be openly shared for AI development. To overcome this hurdle, cooperative consortia could enable secure data sharing, facilitated by deep tech innovation that preserves privacy and safeguards data rights. Such initiatives have the potential to unlock the transformative capabilities of AI in engineering design and manufacturing.
It is an area upon which Digital Catapult and the High Value Manufacturing Catapult are focusing, supported by Innovate UK; projects and programmes often facilitate secure data sharing among notoriously secretive and competitive commercial players, helping to overcome barriers and unlock the potential for deep tech including AI to solve complex market challenges. The parties involved have already started on this journey in the Digital Supply Chain Hub, with virtual testbeds focused on domain specific data sharing in food, hydrogen, textiles and automotive spares.
Culture and strategy
Adopting AI often requires a shift in organisational culture, processes and skill sets. Without proper training, stakeholder buy-in and alignment across departments, manufacturing businesses can struggle to unlock the full potential of AI-generated insights.
This is often due to leadership’s limited understanding of the transformative benefits a well-executed digital and AI strategy can deliver. A clearly defined and well-communicated purpose for incorporating AI into decision-making processes is essential to ensure its success.
For more articles like this, visit our Industrial Data & AI channel.
