Rockwell Combines Generative AI and Industrial Automation

The reliance on text bases in modern software development practices means Rockwell Automation can capitalize on generative AI's ability to read, produce and generate text.

See Also: Trust as the Catalyst: Nurturing Customer Confidence Throughout the Journey

The initial use cases for ChatGPT at Rockwell Automation weren't clear since the large language model-based chatbot isn't designed to write complex code, according to commercial program manager Adam Gregory. But Gregory said the company's new FactoryTalk Design Studio tool brings modern software development practices into industrial automation to fuel collaboration and permit multiple controllers (see: How Microsoft, Rockwell Deploy AI for Faster, Cleaner Design).

"How can we take advantage of generative AI? What can we do with it to make tools for our users?" Gregory said during a Rockwell Automation Fair press conference in Boston. "This is pretty neat stuff, but is it really applicable? We did see a lot of value that we can begin bringing to our users."

Code Snippets and General Q&A

Having open text-based files and a domain-specific language made it very easy for Rockwell to apply generative AI to FactoryTalk Design Studio, Gregory said. A primary use case within Design Studio is "smart creation," or using generative AI to create snippets of code within the automation design software and scaling up from there, according to Gregory.

"What can we do with it to make tools for our users?"
– Adam Gregory, commercial program manager, Rockwell Automation

Gregory said Rockwell wants to provide a paired code programming experience to help people without extensive knowledge or expertise in writing code. Normally, a less-experienced coder would turn to a more experienced colleague for help with programming an industrial pump or valve. Rockwell wants to automate that experience with generative AI, by providing sample code or code snippets to help the programmer get started.

"We've coined this term, 'coder's block,' instead of 'writer's block,'" Gregory said. "If you have one of those problems that you're trying to overcome, you just need a little help in the right direction. One of the initial use cases for us is creating that code snippet."

Gregory said Rockwell has created a general Q&A tool in which the generative AI engine can look through all of the company's product help files to answer questions such as 'What is a smart object?' or 'How do you create a new smart object?' The embedding of generative AI means queries can occur using natural language prompts, meaning coders can get information even if they don't know the exact search term (see: Programming, Self-Learning Crucial for Autonomous Operations).

The responses to queries in the Q&A tool come back in human-readable text, with Rockwell teaching its products how to use domain-specific language and its large language models how to answer questions, according to Gregory. Going forward, he said, Rockwell's products and LLMs will be able to work together to do much bigger and better things.

Generating Code From Customer Libraries Using AI

Gregory said early adopters of Rockwell's generative AI technology want the company to generate not only example code but also snippets directly based on the project content and libraries of a particular customer. Once Rockwell trains its products and the LLM on how to work together, customers will be able to use their own libraries to generate code, according to Gregory.

From there, Gregory said, Rockwell will focus on generating larger groups of project content based on the information users are providing. As humans move from creating projects to maintaining them, they can query the generative AI engine about the purpose of specific routines or programs as well as request assistance with debugging during design time or troubleshooting during runtime, Gregory said.

Regarding documentation, the generative AI engine can describe in human-readable terms what the project is doing as well as provide routine comments and run comments and documents to deliver to the end user and to the OEM. Rockwell worked with Microsoft on the creation of these capabilities and uses Open AI to communicate with Rockwell products running in Azure, he said (see: Rockwell Forges Gen AI Pact With Microsoft, Buys Cyber Firm).

"Rockwell knows the OT side of things. We're bringing that understanding of how to build and how to program pumps and valves," Gregory said. "And then Microsoft is helping us with the cloud side of things and also then with analyzing that data and how to make our prompting work better."