Since IBM’s Watson, many have hoped that Artificial Intelligence, or AI, would be able to rummage through a pile of data, pick out what is relevant, and create some new learning that matters. That hope has been proven futile, at least with our current thinking and capabilities.
That doesn’t mean AI is useless; we have learned through trial and error that one has to have a clear understanding of the question he is asking AI to answer. The more narrowly we define the question, and the more relevant the data fed to AI, the better it will uncover an insight of value to you and your business.
Perhaps it makes sense for you to begin by working to understand key contributors to variation in scrap of a particular type of metal that you process, or on certain equipment, or level of product design complexity. AI needs accurate and relevant data inputs over a period of time with sufficient detail to provide valid output.
AI might discover that ambient temperature impacts scrap, but only if that is part of the data at a sufficient level of detail. The same is true for die temperature or pressure variations. AI will produce invalid or inadequate output if the input data doesn’t include potential contributors to the problem you want to solve.
Once you have gained experience with AI, you will learn how to better examine broader and more lasting questions.
AI is NOT a starting point of Industry 4.0, nor for most of you is it even early stage. Processes to provide clean data, some of which sensors can generate, but some of which will come from other systems or from people, must be mastered for the technology to help you solve big problems.
Podcast: Download