Issue link: https://iconnect007.uberflip.com/i/1535954
JUNE 2025 I SMT007 MAGAZINE 65 but what you're doing isn't necessarily new. Where else has it gotten traction? We have served the consumer and auto- motive markets, where this is becoming more commonplace. You can ask questions about specific products, about specific car models, for instance, using an AI large lan- guage model that is pre-loaded with the information that pertains to this item. e AI we use started as a model that augments a large language model. We call it a retrieval augmented generation in the industry. It was very rudimentary in the beginning, and now we are evolving into what we call multimodal information, where we add voice, images, and videos in addition to basic text, which is what large language models are primarily built on. With all those different modes, you can com- municate with an AI system, and eventually it will go out into augmented reality as well. We use classes for maintenance operations and so on. ere are multiple ways to interact with the system, and you choose for yourself which is the most appropriate. As this approach becomes more widely adopted in our industry, what do you envision for the future? You will have a much-improved way of operat- ing machinery. You can do changeovers much quicker than you do right now. You can adapt to new pieces of equipment—or new combina- tions of equipment—much faster. You'll be up and running much quicker. You will probably have fewer operators, perhaps even dark manu- facturing. To keep the machines running, oper- ators will need much less training time. If any- thing happens, the incidence rate of anything going wrong will be dramatically reduced. But if something does happen, the AI will also quickly learn from that experience. Proactive mainte- nance is much more common, but will proac- tively suggest things you can do to prevent a problem. We will have so much more uptime and a much more automated production pro- cess than we have right now. at's what I see happening in the industry. At IPC, the CFX inter-machine communication protocol is based on some banking proto- cols. So, there is some security to it, allowing machinery in the line to communicate with each other. is opens the possibility of keeping things better centered in the process window. If this machine finds that the process is on the edge, then perhaps some process further down can adjust that board to pull it back toward the cen- ter. at's the big picture: better process con- trol with the machines themselves keeping the process from spinning off. at information is stored in local databases at the EMS com- pany; they're creating their own company-spe- cific big data. It seems there's an opportunity to connect with all this process data using your point of entry into your AI. You could provide access to the process con- trol panel, if you will. Operators could walk by the machine, tap, and check compliance with the process window based on historical data. In the future, there will be a lot more interac- tion. Data exchanges like CFX or Hermes will