Making your machine talk -- how generative AI could transform the packaging and processing machinery industry
For CPG customers struggling with declining availability and quality of workforce, OEMs have a unique opportunity to leverage generative AI to solve real customer problems.
At PMMI’s recent Road Show in July in Vancouver, WA, keynote speaker Shawn French, who holds a leadership position in packaging engineering and innovation at Danone, spoke about a critical technological advance early in his career: the introduction of HMI’s.
Prior to a visual display, he recalled technicians spending anywhere from 10 minutes to 10 hours trying to troubleshoot a machine when it went down. Having a window into the machine’s statuses was revolutionary, and contributed significantly to reducing downtime.
Another data point: At PMMI’s Top to Top meeting earlier this year in March, one of the topics that surfaced was a debate around guidance versus training. Training made more sense in an era where employees had greater aptitude for operating equipment and stuck around longer. With many CPGs struggling with higher turnover and lower workforce quality, an increased emphasis on in-the-moment guidance on how to do or fix something seems to make more sense versus training, which loses its efficacy with a revolving-door workforce.
At that same meeting, when CPGs were asked about their top needs OEMs should take into account when buying packaging equipment, workforce jumped from number seven in 2023 to number three on the list this year, after only productivity and automation, two perennial factors that don't seem to change."Workforce" jumped from number seven to number three on the list of CPG priorities in just 12 months.
As Shawn French said at the Road Show, machines need to operate more intuitively. “Do you remember your iPhone training?” he asked the audience. It was a tongue-in-cheek reference to the fact that the groundbreaking user interface, which today we all take for granted, was so intuitive at the time that no training was needed to operate what is essentially a highly sophisticated pocket-size computer.
What do HMI’s, the guidance-versus-training discussion, and the declining availability and quality of workforce have to do with AI?
In November 2022, if you had asked me how close we are to an artificial intelligence with which you could have actual conversations, I would have scoffed and said that’s the stuff of science fiction and that we’re at least 50 years away. The following month I, along with the rest of the world, found out about ChatGPT.
We have all watched with amazement (and not a little trepidation) at how quickly this technology is developing. I’ve often wondered how packaging and processing OEMs might leverage these technologies for their equipment. Surely the big automation suppliers (Rockwell Automation, Siemens, Beckhoff, B&R, etc.) are looking closely at how generative AI can be incorporated (safely) into their platforms. Our editors will follow and report on those developments. But that will take time.
In the meantime, there is room for adventurous OEM engineers to tinker with generative AI directly, potentially making their machines much smarter and much more forgiving. The goal: lower the demands on operators and technicians at their CPG customers to successfully operate the equipment. More on that shortly.
But first, who am I, and what qualifies me to hold forth on this subject? I am neither an engineer, nor an AI scientist. But I’ve worked in the packaging trade press for 30+ years, first covering packaging machinery as a journalist and subsequently working my way through various roles at Packaging World Magazine and its parent company, which was eventually acquired by PMMI. (I currently serve as PMMI Media Group’s president.)Dave Newcorn, guest author of this column, owing to the flimsy confluence of having dabbled in arcane-and-useless programming languages 40 years ago and having covered the packaging machinery industry for 30+ years. Day job: PMMI Media Group president.
I’m also an ex- (or failed) programmer, having cut my teeth coding in assembly language in the early 1980s, a low-level programming language perhaps at least as arcane as ladder logic. The admittedly flimsy confluence of these two things affords me just enough perspective to prognosticate dangerously, which I will now proceed to do. Please fasten your seatbelt.
One final note: Sean Riley, OEM Magazine’s esteemed editor, has graciously agreed to let me guest-write this column until we get somebody better. In the meantime, you are stuck with me. Let’s get started!
Your machines, talking back
It’s only a matter of time before generative AI is baked into every consumer product, starting with our cars. While vehicles have had voice interfaces for years now, they are crude and semi-functional at best, especially compared to what we’ve come to expect from ChatGPT and its ilk. Soon cars will start showing up with their own ChatGPT style chat interfaces, allowing drivers to ask how to access or activate a given feature, or provide much clearer information on a problem that a vehicle is experiencing. Quite frankly, I expect any consumer product that is remotely technical will come with its own generative AI chat interface—smart TVs, home security systems, etc. There’s just no reason not to.
Prognostication number one: So it goes with packaging and processing equipment. Packaging and processing OEMs face an unprecedented opportunity to build customized language models into their equipment, amounting to an interactive knowledge base, in the form of a chatbot, specific to each machine. In my view, it’s a shift no less significant than the one from mechanical motion control to servos.
Imagine an operator pressing a push-to-talk button on your machine and speaking – in any language – to ask about a particular setting that might be unfamiliar, how to troubleshoot something, or even to ask the machine to double-check a particular setting after a changeover. For shift changes where the equipment is operating inconsistently vs a previous shift, imagine a technician or supervisor asking the machine to review and compare the current shift’s production data with that of the previous shift and to point out any anomalies in how the machine is being operated, changes in inputs, etc.
(You can do a crude version of this right now by simply uploading a spreadsheet of production data to the paid version of ChatGPT. Once you contextualize the data and explain what the various columns mean, you can ask ChatGPT to identify trends or patterns in the data, even if that data consists of hundreds or even thousands of rows.) Endowed with a generative AI interface, your machine can now speak back, intelligently. In any language.
This has the potential to significantly compensate for the degradation in workforce quality and continuity that today’s CPG customers are struggling with. Suddenly operators don’t have to be as knowledgeable to successfully operate a given machine. The days of your HMI spitting out inscrutable fault codes will come to a close. Operator and technician training burdens have the potential to be significantly reduced. If we get the technology right, it’s potentially game changing for machine builders as well as for your CPG customers.
We have all heard about Large Language Models, which, combined with massive quantities of computing power, enable the phenomenon of generative Artificial Intelligence. In my next column, I will drill down on the lesser-known concept of Small Language Models, which are a fraction of the size, can enable the functionality discussed in this column, and which can run on a PC embedded right in your machine.
OEM Magazine is pleased to inaugurate this semi-occasional column tracking the rapid advances in AI and how packaging and processing machine builders can leverage them to build next-generation equipment. Reach out to Dave at [email protected] and let him know what you think or what you’re working on when it comes to AI.
"Workforce" jumped from number seven to number three on the list of CPG priorities in just 12 months.
