Vision for machine learning in optical inspection
DarwinAI develops machine learning and artificial intelligence solutions to improve AOI inspection results. Their work is unique as it is independent of the hardware typically used in EMS lines. It represents a pure solution using machine learning and AI constructs wherever possible in our industry.
Earlier this year, I-Connect007 staff spoke with DarwinAI CEO Sheldon Fernandez about artificial intelligence and electronics manufacturing. In this follow-up, Sheldon explained what AI and machine learning are, what they aren’t, and how this effort fits into today’s manufacturing environment.
Nolan Johnson: How will artificial intelligence affect industrial environments like electronics manufacturing?
Sheldon Fernandez: Emphasis was placed on visual inspection. Artificial intelligence step-change capabilities to perform such tests are becoming important in the industry. Here we focus on the last third of the workflow, the backend assembly after SMT. This represents an underserved aspect of the PCBA process. This is also an obvious area where deep learning and artificial intelligence will see immediate productivity gains.
Of course, it would be irresponsible not to mention the explosion of the AI space with the generative AI that ChatGPT is based on. This is when AI produces assets that are useful to humans in all situations, such as natural language, images, computer code generation, and script code. We’ve already seen this about coding, but let’s think about the possibilities of generative AI for PCB designers. I think designers, as well as people who create design tools, can leverage this technology in productive ways. Just to be clear, it’s not perfect. But if you can get to 70%, you can do the rest. Those are the two areas that immediately come to mind as someone in the AI field.
Johnson: Design-wise, we’ve been using automatic routers for a long time. In semiconductors, 25-30 years ago we had IC chip compiler tools. Will those tools be called AI?
Fernandes: Maybe so. What do we mean when we say AI? The term has now come to mean anything that produces a useful asset or insight. AI experts usually think of AI as the second or third wave of machine learning and deep learning, aspects of artificial intelligence that make use of complex models that analyze and learn from vast amounts of data. thinking about. No doubt some of the tools you describe have done much of this work. Now, I’m no design expert, but what is the productivity step-change? For example, in software development there are always tools to speed up debugging and help with code generation. But I don’t have a tool that gives me Python, C-Sharp, or Java code to say, “This time, with the idiosyncrasies of this algorithm, write an algorithm that sorts a million numbers by these properties.” was. It’s pretty amazing how easy it is for a non-technical person to achieve such results. When I refer to AI, it’s the sophistication factor.
To read this full conversation published in the July 2023 issue of SMT007 Magazine, click here.
