AI in industrial automation: from chatbots to plant chums and hallucinations

With the current hype around AI, its impact on instrumentation, control and industrial automation must be considered. Software companies that reference it are certainly enjoying an upward blip in share price; but is it really that useful?

Essentially, the AI algorithm is based on a vast repository of content extracted from the web and underpinned by computing power.

In November 2022, OpenAI announced the Large Language Model (LLM) and associated ChatGPT — an interactive ‘chat’ bot. Graphics, avatars and songs are also possible with various associated apps. Recently, I was privy to a demonstration of AI technology involving a suitably attired AI-driven professor. She was teaching statistics (logistic regression) and interacting with students very ably. In a similar vein, and with some success, we have trained up an AI-based chatbot to interact with students working on our online industrial automation labs (without divulging the solutions of course).

The tussle between Google and Microsoft and their AI offerings continues; this competition will result in comprehensive improvements.

ChatGPT provides the following note on AI’s impact on industrial automation. It is fairly predictable and prosaic:

The greatest impact of AI on industrial automation is the ability to optimize and improve processes through data analysis and machine learning. This results in increased efficiency, reduced downtime, and improved product quality. Additionally, AI has enabled predictive maintenance, allowing companies to prevent equipment failures before they occur, saving time and money.

To apply AI easily and effectively, users should start by identifying the specific pain points in their workflow that can be addressed by AI. They should then evaluate the available AI technologies and select the ones that best meet their needs. Once the appropriate AI technologies are identified, users can work with vendors or consultants to integrate them into their existing systems. Additionally, users can leverage AI development platforms that allow them to easily build and deploy custom AI solutions without needing extensive programming skills. Lastly, users should prioritize ongoing training and upskilling to stay up-to-date with the latest AI technologies and techniques.

After being suitably impressed with ChatGPT when querying process control tuning rules, dismay followed when information on recent industrial automation advances was requested. This is not surprising; the knowledge store on OpenAI is lagging by about two years.

The immediate value in AI tools is the automation of mundane, repetitive tasks: writing emails and reports, for example. It can act as an office chum: providing a starting point, before a necessary content rework to ensure it is applicable to the plant.

Despite the inevitable disappointments, innovation has improved life and business through time: steam during the industrial revolution; electrical engineering in the 1900s; electronics in the 1960s; and the internet in the nineties. AI should see the same positive growth in industrial automation plants in due course. There is a caveat, however: the overriding need for safety will ensure that the uptake is cautious; for instance, ChatGPT could ‘hallucinate’ (the actual term) while transferring sulfuric acid.

A possible way forward for you:

• Install ChatGPT.
• Investigate the Microsoft and Google options.
• Use Prompt Engineering: to evince accurate answers.
• Use ChatGPT for AI queries and for communication, but be critical!
   

Finally, the spread of AI may also be impeded by some legal kick-back. After all, the current model relies on harvesting data without any recompense to the originators of the content.

*Dr Steve Mackay has worked in engineering throughout Australia, Europe, Africa and North America for the past 40 years in the mining, oil and gas, and power industries. A registered professional engineer in electrical, mechanical and chemical engineering, he believes university engineering programs need to be strongly focused on industry. He has been the author or editor of over 30 engineering textbooks sold throughout the world.

Image: iStock.com/Supatman