Just like the proverbial “kid with a new toy,” everyone is playing with generative artificial intelligence (AI) and ChatGPT to see how their applications and businesses can take advantage of it.
“We’re bringing generative AI to process automation via FactoryTalk Design Studio, and developing case studies we want to do because this is still a prototype solution,” said Adam Gregory, platform manager for FactoryTalk Design Studio at Rockwell Automation, during a press conference at Automation Fair 2023 this week at the Boston Convention and Exhibition Center. “I began learning about generative AI last year and started by asking it to build images and write simple code. We also saw many opportunities when we looked at the GitHub copilot beta and the introduction of OpenAI’s ChatGPT in November 2022. Rockwell Automation builds tools, so we wanted to know if generative AI could help.”
Released in October 2022, FactoryTalk Design Studio is Rockwell’s automation design solution that also lets users simplify and streamline their collaboration and productivity. It’s a natural platform for exploring and experimenting with generative AI and ChatGPT because it’s a cloud-native solution for building control and related software.
Snippets and queries
Gregory reports the Rockwell’s initial goal for generative AI and ChatGPT was just to have it generate some small snippets of code, and then query it about the best way to produce code for a device, such as a pump. Next, Gregory and his colleagues did some general question-and-answers, such as one might ask a coworker, and did it with ChatGPT’s natural-language prompt.
“This allowed us to create a foundation that we could build on and iterate with,” explained Gregory. “We use the code snippets and question-and-answer results to gather more input, which could be used to do bigger and better tasks, such as automated creation actions based on each user’s data and libraries. It could also tell us what a project was doing, help with related tasks, inform us about what’s routine or not, and run comments from other users.”
Gregory added that Rockwell’s FactoryTalk Design Studio team collaborated with Microsoft Azure’s OpenAI design service, and got help with cloud-computing, analytics functions and improving prompting.
“This is still a prototype, so it’s not available to users yet,” explained Gregory. “But we did develop a FactoryTalk Design Studio copilot, which uses the natural-language queries. This is possible because all of ChatGPT sits on large-language models (LLM). We’re now trying to incorporate Rockwell Automation’s domain know-how with them.”
Code and alternative code
Despite some initial hiccups, Gregory also reported that generative AI and ChatGPT successfully produced the code snippets his team was seeking about how to run a pump, and even accommodated additional user requests.
“The initial snippet recommended using latching and unlatching in ladder logic to control the pump, but we decided we didn’t want to do that, and asked if there was different code for another option,” added Gregory. “The copilot removed the code we didn’t like and suggested using an output-energized method that it subsequently added. This means users can keep working with this software to find new options and alternatives. It also means we can teach models how to do automation by defining engineering prompts and sending them to a static LLM. It really is a copilot, and not an auto pilot.”
