Machine Learning: What, Why, Where, How and When

Analytics Working Group, MESA Point of View (PoV)

What does machine learning mean to you?  A popular notion today is that in the very near future we will be working side-by-side with autonomous, intelligent robots as collaborative colleagues.  Or perhaps when you think of intelligent machines, you dream up lawless replicants being chased around the galaxy by Harrison Ford.   If these are the things you conjure up when you hear the term machine intelligence, you are not alone.   There is confusion in industry as to what machine intelligence and machine learning are and how they will affect and benefit manufacturing industries.

What?

Machine Intelligence today is an application of algorithms and analytical techniques to allow computers to leverage data more effectively. This is a foundational technology for big data analytics – being able to parse data into more meaningful sets of information.   Machine learning is a neural network approach that allows machines to learn without humans programming all the rules. This means they get better at their tasks as they perform them.

Why?

When machine learning and intelligence are built into production machines, they can operate more efficiently, accurately and safely.  Fortunately, machine learning and intelligence are not futuristic or fantastically expensive technologies that are out of reach.  Rather we can and should be applying analytics today orchestrate our machines and processes, maximizing each asset’s capability. 

Where?

The most near-term evolution of this technology would be to apply analytics and governing algorithms to coordinate activities between machines.  Think self-assembling, flexible processes, intelligent safety interlocking, proactive security and predictive maintenance. Many companies have maintenance applications using machine learning already.

How?

To take advantage of opportunities to apply analytic processes and create the closed-loop feedback required to raise the intelligence of your machines,  we need to prepare our organizations.  We need to identify the sources of data and adopt standardized integration techniques to expose and collect that data.  We need to adopt and enforce an enterprise framework, based on open standards which will provide the appropriate context for this data.  We need to be breaking down barriers between departments, customers and suppliers to allow for more collaboration. As we expose richer data sets, machine learning and machine intelligence gain value.

When?

Clearly the driving forces for manufactures remain consistency, efficiency, flexibility and high quality. Autonomous intelligent machines that learn as they go might assist that, and in some cases already are. That said, for most companies the next step in the natural evolution of machine intelligence will not be robots talking interactively with humans, collaborating on supply chain decisions. Rather humans can apply analytics and business know-how to make the most effective use of the assets they currently possess. This is not the future, this is best practice today.