From large language models to synthetic data generation, AI is analyzing, organizing and creating the data that drives digital transformation for engineers and manufacturers.
Ever since OpenAI launched ChatGPT in November last year, it seems as if the world has turned on its head. The large language model (LLM) has managed to reignite debate over the future of work, media, entertainment and even mankind itself—but for engineers and manufacturers, artificial intelligence (AI) promises to be so much more than a text-generation tool.
As organizations shape their transformations towards their desired outcomes, AI’s role in collecting and analyzing data will be key. This is already being seen in the development of chatbots, for example, as organizations explore opportunities for using LLMs to improve workflows for designers, or even in knowledge transfer. We are also seeing it with synthetic data, algorithm-generated data designed to help train AI models and create targeted 3D simulations.
AI is all about data
To accelerate modelling and improve accuracy, “fake” photorealistic data can be used to build reliable, accurate and optimized machine learning models. In a July 2021 blog post, Gartner shared its prediction that 60 percent of the data used for developing AI and analytics will be artificially produced by 2024. The reason is that taking a scatter-gun approach to data collection is inefficient and fraught with complications in terms of data quality, accuracy, bias and relevance. It’s also costly.
Digital tools that enable the use of synthetic data, that can train and deliver AI capabilities, will increasingly become the norm, but for the moment they are a rarity. This means they currently represent a competitive advantage, and the use cases are growing. For example, companies like Siemens are using NVIDIA’s Omniverse platform to generate synthetic data to train models for machine defect detection, robotic bin picking and safety monitoring. An NVIDIA blog post from November 2022 claims that Siemens has already turbocharged its AI development times, reducing it from months to days.
It’s a good example of how an organization is positioning itself to take advantage of the latest data and AI capabilities. As Accenture claimed in a recent report, Is your organization equipped for breakthrough innovation?, cloud, AI and the metaverse “are accelerating into megatrends that have the potential to dramatically speed up the pace of technological change, bend the innovation curve and become a crucial part of every organization’s value chain.” It is that speed of change that can be unnerving, so organizations have to look at technologies and figure out which is going to help them improve their operations, innovation and decision making.
AI undoubtedly is going to be the driving force of automation and is expected to underpin software and services through advanced analytics and interactions. As Mike Glennon, IDC senior market research analyst, recently commented in an IDC press release, “companies that are slow to adopt AI will be left behind—large and small. AI is best used in these companies to augment human abilities, automate repetitive tasks, provide personalized recommendations and make data-driven decisions with speed and accuracy.”
AI makes smarter humans
Another area where organizations are going to see increased activity with AI is knowledge transfer and acquisition. Smart human and machine interactions will be an important focus for industry in the coming years and building knowledge both for machines and humans will be key.
These smart interactions can enhance and enrich the mechanical design process, with AI acting as the linchpin, both in terms of relevant data collation and analytics and also in terms of prioritization. As Shirish More, product manager for Siemens NX suggested in a June 2022 blog post, “Artificial intelligence—especially when it comes to mechanical engineering software—is our ability to recognize design patterns and solutions.”
Uncovering patterns across a wide array of sources, as well as sharing and recommending workflows based on accumulated knowledge, is part of the decision-making process. AI leverages the wealth of engineering knowledge in completed designs, to support the sharing and transfer of knowledge within design groups and between people. This knowledge transfer, traditionally achieved through a mix of manual processes and data-driven insights, will drive learning across organizations—but it has to be relevant.
This is where AI comes in, generating models based on organizational need. To a certain extent, firms are seeing this with generative AI. ChatGPT, for example, can help support engineers through advanced pattern recognition and workflow optimization abilities, as well as personalized training and identifying career development opportunities. The ability to use AI to predictively improve workflows, suggesting best practices and design choices along the way, can also boost knowledge and supplement skills gaps.
The ability to analyze patterns and predict outcomes is not new. Analytics derived from multiple sources—shop floor data, logistics, sales, customer insights (buying trends, regions, etc.), product maintenance data—is already being used to optimize maintenance teams as well as research and development teams. But AI adds a new level of organization and insight. With AI constantly analyzing data and learning patterns, adding new sources and growing its picture of an organization, design models, products and structures, the more it adds value to decision-making, ultimately driving the whole product lifecycle management (PLM) process.
The ability to answer the “what if?” questions means organizations can plan intelligently, based on predictive models—predicting rather than just reporting. This filters across an entire organization. With this level of data and analytics, organizations can begin to optimize functions, such as supply chains, finance, sales and product design. It can even help with personalization, especially in skills development and user experience.
The lifeblood of innovation
As organizations embark on digital transformations, access to data will be the lifeblood of innovation and production, but the challenge is to prioritize. While digital twins, for example, may be an ultimate goal, understanding the role of data and AI in making sense of that data is fundamental to planning and delivering accurate and relevant models. As AI has moved out of the labs and into practical applications within engineering and manufacturing, it is the machine’s ability to deliver nuanced intelligence that will inform decision making through these models.
This is the building block on which all future engineering and manufacturing firms will be built. Interestingly, IDC claims that manufacturing (both process and discrete) will account for 16.6 percent of the AI market, by far the biggest individual industry sector. Much of this is driven by manufacturing’s long-standing relationship with automation. The drive for autonomous factories continues, but unlike previous iterations of automation, AI is already very different. It won’t be too long before decision-makers will be looking back and wonder how on earth they managed without it.
