It was Aristotle who said the whole is greater than the sum of its parts. Systems thinkers believe this concept to be particularly true; they find a ‘wholeness’ in systems. Machine learning is a system in and of itself. When analyzing the particulars of a system using the machine learning process it is best to understand a system is being used to study another system.
Machine learning models are the end result of a system. The model is the output of the elements, interconnections and purpose that exists within the machine learning system. The elements of a system are typically the easiest to notice. In machine learning the elements could best be described as the raw, unstructured data that are brought to the engagement. The interconnections in machine learning are the physical flows and reactions that are associated with the system’s process – prompts that have one part of a system responding to what is happening in another part of it. Obviously, interconnections in a system are much harder to notice than elements, yet the interconnections in the system are precisely what we are searching for. In machine learning the enrichment of data through labeling and annotation will have a tremendous affect on a system’s elements and to a large extent its interconnections.
While purpose in a system is unquestionably the hardest to spot and comprehend, purpose cannot be discovered without a thorough analysis of the elements and interconnections inside the system. That’s what makes machine learning such an important part of business today. To argue that purpose is the most important part of a system or model is to take an ‘unsystematic’ view of an engagement. Elements, interconnections and purpose are all essential in a system or model. They all interact, and each has its role.
The complexity associated with machine learning makes clear that no part — and no process — is more important than another. Further, it strongly suggests that the whole is always more than the sum of its co-equal parts. Understanding the differences in elements, the quality of relationships and interactions, and the actions and reactions that lead to both predictable and unpredictable outcomes is the benefit of machine learning. Innovative business leaders who are looking to machine learning to help plot a course for corporations in the Age of AI will need to invest in data, data labeling and annotation, and the building of machine learning algorithms and models before they can reap machine learning benefits. — Mark Papia
(Mark Papia is the Director of Content and Communications at iMerit.)
