Issue link: https://iconnect007.uberflip.com/i/1528798
NOVEMBER 2024 I SMT007 MAGAZINE 13 model on the production floor just on the factory performance data? Burke: I don't want to get hung up too much on LLM vs. SLM in terms of size of the model; it's not as critical. What matters is the intention of what that thing is doing. It's making a live deci- sion on the shop floor in support of a supervi- sory role or in support of a process engineer trying to get to the root cause of the event— perhaps in support of a quality engineer trying to decide whether this particular work order is high risk and in need of an audit, or actually based on the collection of data I have, whether I should just ship it out the door knowing it'll be totally fine. Both the large and small versions of language models have enough capability now to actively help make those decisions, which opens a whole set of use cases that previously weren't considered possible. This is indeed a different type of data science expertise. You're taking the data you're get- ting out of CFX, for example, integrating it into your MES and making dashboards like you've never seen before. Burke: Yes. Previously, we thought the process was like this: Get a bunch of data, hire a data scientist who does something hard with the data ("data science"), and turn it into a really