PHYSICS, MACHINES AND DATA

Machine learning is a mathematical construct that is the foundation for nearly all the advancements in AI. ML came first, but it remains relevant even today. It can applied to semiconductor fab for such things as predictive maintenance of manufacturing equipment, rather than just maintenance on a schedule, which decreases downtime. But getting this right is harder than it sounds. Data needs to be relevant, clean, and organized in the right form. Jon Herlocker, VP & General Manager of Tignis, A Cohu Analytics Solution, talks with Semiconductor Engineering about what can go wrong in gathering and applying data, why so much compute horsepower is required, and where to harvest the data needed to keep it relevant. This is the second part in a seven-part series on AI in chip manufacturing.