Generative AI is hailed as a revolutionary force in manufacturing. But, to truly transform factory performance, we must understand both its capabilities and limitations. Otherwise, we risk stifling AI’s potential rather than unleashing it.
It’s time to separate Gen AI hype from reality, says XpertRule CEO and Founder Akeel Attar.
The buzz surrounding Generative AI is at a fever pitch, with many touting its potential to revolutionize manufacturing. Yet, having been immersed in AI technologies for the past four decades, I’ve observed the pitfalls of overpromising and the disillusionment that follows when technology fails to meet overblown promises.
If you take nothing else from this article, remember these core realities:
- No single solution can address all business problems
- AI is not one technology, it’s a collection of different tools and techniques
- Matching the right AI technology or combination to each use case is essential
- Generative AI excels in language tasks but is less effective with numerical data
What is Generative AI?
For many, Generative AI burst onto the scene two years ago with the release of tools like ChatGPT, introducing many people to large language models (LLMs).
LLMs are trained on vast amounts of data to process natural language and create human-like text. This makes them highly effective for language-based tasks such as generating content, answering questions and processing natural language inputs or prompts. Gen AI capabilities extend beyond text to encompass new content across various forms, from images to designs and even music.
As with any groundbreaking technology, it wasn’t long before bold claims emerged suggesting that Gen AI would transform every industry, from agriculture to space tourism. Manufacturing, with its emphasis on efficiency, precision and quality control, appears to be an ideal candidate for leveraging Gen AI.
However, it’s important to approach these claims with a critical eye. It is time to step back and temper enthusiasm with a clear understanding of where Gen AI can truly add value and where its limitations lie. The real question for manufacturers is: How can these capabilities best be applied in an industry deeply rooted in numerical precision and data analysis?
Gen AI in Manufacturing
Read any article discussing Generative AI in manufacturing and you’ll likely see recommendations for predictive analytics and maintenance, condition monitoring, process control and optimization as use cases. This is to be expected, given that these are all major focus areas for manufacturers, as they directly influence efficiency, productivity and cost-effectiveness.
However, in suggesting these applications, the writers overlook a clear limitation of Gen AI. Each of these use cases relies heavily on numerical data – sensor readings, performance metrics and quality measurements – something Gen AI, primarily designed for text-based tasks, is poorly suited for.
The key issue lies in how Gen AI models, such as LLMS, are structured. These models are very good at understanding and generating text because they are trained on vast amounts of unstructured linguistic data. However, when it comes to interpreting and analyzing structured numerical data, such as time series data from machine sensors, Gen AI lacks the precision and consistency needed for generating reliable, predictive and optimized outcomes. Furthermore, Gen AI is unable to automate expertise when there is insufficient data to model outcomes.
This is where the adage “if all you have is a hammer, everything looks like a nail” serves as a cautionary reminder. Just because Generative AI is a powerful tool doesn’t mean it’s the right tool for every problem. Forcing Gen AI to perform tasks it’s ill-suited for can lead to inefficiencies and unreliable results – this can be disastrous, especially in manufacturing where precision and consistency are critical.
It is especially frustrating to see Gen AI put forward as the solution in scenarios where data analysis tools, rule-based systems or human expertise would be more effective. Machine learning (ML), for example, is a proven technology that has been a staple in manufacturing for years. ML algorithms are specifically designed to recognize and analyze patterns within numerical frameworks. They leverage regression analysis, classification and clustering to identify understandable patterns, making them far more suitable for tasks like condition monitoring and process optimization.
Applications of Gen AI in Manufacturing
The strength of Generative AI lies in areas where understanding and generating natural language is critical and is highly adept at extracting information from unstructured data, like documents, emails, audio files or videos.
While Gen AI may not be the silver bullet for all manufacturing challenges, it does offer value in specific areas. Over the years, my engineers and I have successfully applied Generative AI across several practical applications:
- Document Analysis
Gen AI can streamline extracting key information from documents, such as contracts, invoices, maintenance logs and compliance records. For instance, it can quickly extract terms, conditions and deadlines from supplier contracts, reducing manual effort and the risk of errors.
- Customer Support
Gen AI-powered chatbots or virtual assistants can generate automated responses to common customer queries about product specifications, troubleshooting steps and order status, freeing human agents to handle more complex issues.
- Internal Communication
Generative AI can assist in drafting documents, creating standard operating procedures and summarizing reports, allowing employees to focus on strategic decision-making rather than getting bogged down in routine administrative tasks.
- Voice Control
Voice-controlled interfaces powered by Gen AI can allow operators to issue commands or retrieve information without interrupting their work, improving safety and productivity on the factory floor. A maintenance operator, for example, could use voice commands to receive the next step in a repair procedure while keeping their hands on the equipment.
- Data Preparation
High-quality data is essential for training machine learning models. Generative AI can generate scripts for cleansing, formatting and enriching information from different sources, ensuring that ML models receive accurate and well-structured data to enhance their predictive capabilities.
Navigating the Hype
For decades, AI has been overlooked, with many industries slow to leverage its capabilities. Now, with the sudden surge of interest in AI (largely driven by a ‘fear of missing out’) it’s vital not to get swept away by the hype and lose sight of the bigger picture. While the excitement around Gen AI is understandable, we should approach its implementation in manufacturing with caution.
To truly unlock AI’s transformative potential, manufacturers must understand where and how to apply these technologies effectively, and not ignore their practical limitations. By understanding the distinct roles of Gen AI, Predictive AI (ML), optimization techniques, and rules-based AI and using them in combination, manufacturers can avoid the pitfalls of overhype and instead achieve meaningful, sustainable improvements in factory performance.
This article is the first in our new series, Reality Check: What AI Really Means for Manufacturing, designed to inform, inspire and help you implement AI in your manufacturing operation.
Look out for our next article on AI in the Pharma sector ahead of PharmaTech Integrates 2024 on 19th September. XpertRule’s Iain Crosley will be on stage discussing practical applications of AI in Pharma and Life Sciences.
For more info and to register, please click here
Sep 5, 2024 10:39:21 AM