Innovation and Technology

A common theme in science fiction is the fully automated, robotized factory that manufactures nothing but robots. We’re not there yet, but the fully automated manufacturing plant has already begun making everyday products, including computer parts, electric shavers and CNC machines.

The promise of AI: Now generative AI is promising to take manufacturing automation manufacturing to a new level.

• At the 2023 Hannover Messe trade fair in Hanover, Germany, Siemens and Microsoft showcased an offering now in use in factories worldwide: a system that uses ChatGPT to generate code for industrial computers known as programmable logic controllers. (For a deeper dive into what this means for manufacturing, read the full version of this article by Tim Hornyak in the Innovation Research Interchange’s Research-Technology Management magazine.)

Why it’s important: The innovation allows users to ask ChatGPT to generate code for specific tasks (i.e., a program to operate the stamping of a part).

• In addition to saving time and reducing the likelihood of errors, it is capable of understanding commands given in natural language, a characteristic that vastly increases the number of potential users.

Efficient designs: Creating more efficient designs is another early use case for generative AI.

• General Motors has used the technology to evaluate better designs for some of the roughly 30,000 parts that go into the average car. For example, a standard seat bracket—an important safety component that binds seatbelt fasteners to seats as well as seats to the floor of the car—consists of eight separate pieces welded together.
• Generative-design software used by GM analyzed the requirements and suggested more than 150 alternative designs, far more than the two or three options a designer can typically offer. GM engineers chose one: a single piece of stainless steel that is 40% lighter and 20% stronger than the conventional part.

Pharma applications: Generative AI looks promising for the pharmaceutical industry, too, given its potential for cutting costs and drug time to market.

• Merck has used generative AI to create synthetic images of complex but rare defects, a group for which training data are limited. The drugmaker’s quality-control sensors use the synthetic images to watch for novel defects.

Other potential use cases: The possible uses for generative AI in other areas is vast and includes the following:

• Reducing time and cost involved in creating physical prototypes
• Automating search and summary of documents related to manufacturing equipment, which would speed repairs and maintenance
• Accelerating supply chain operations by forecasting demand patterns, minimizing production downtime and suggesting better transport routes
• Customizing products or solutions to better suit customer needs
• Forecasting raw materials needs, optimizing production schedules and identifying production inefficiencies

However … Generative AI in manufacturing is not without its challenges. The energy cost to power a single server rack in the U.S. is $30,000 a year.

• Just one training run for an AI engine consumes the power equivalent of 120 U.S. households per year.
• With the reliance on large datasets, manufacturers are concerned about data privacy and security, necessitating robust data-protection measures.
• The integration of AI in manufacturing may require a change in workers’ skillsets and corporate culture.
• As AI plays a more significant role in decision-making, ethical questions about bias and accountability are emerging.
• Manufacturers have to ensure that AI systems operate fairly and transparently.

Find out more: AI in manufacturing is just one of the timely topics covered in depth in Research-Technology Management, the journal of the IRI, the NAM’s innovation division. Learn more.