As an IT Solution Architect, I am skeptical of any over-hyped technologies or so-called silver bullets. Recent buzz in the IT press indicates that Cognitive Computing belongs to this category. As a practitioner in the previous generation of Artificial Intelligence (AI), I know what happened after the over-hype … a trough of disillusionment. If we want to avoid another AI Winter, we should heed the lessons from the previous AI era. Some of these lessons are to avoid the hype, manage expectations and ensure sustainable deployment.

This article tries to sidestep the hype and to uncover what is Cognitive Computing from a practitioner’s point of view and how it differs from the previous generation of AI. The focus is not on the theoretical aspects AI but on the practical perspective required to apply Cognitive Computing on real-life problems.

Definition and differences from previous AI

Cognitive Computing is defined by some as “the simulation of human thought processes in a computerised model”. Such definitions worry me because they sound very similar to statements from the first generation of AI and do not tell us much. Other definitions progress our understanding a bit (but, unfortunately not far enough). They define Cognitive Computing as “systems that learn at scale, reason with purpose and interact with humans naturally”. Yet other definitions add Adaptive, Interactive, Iterative and Contextual to this list. These characteristics, no doubt, are highly desirable but do not tell us how Cognitive Computing achieves them. They do not help in deciding if we should use Cognitive Computing to solve a given problem or how we go about architecting the solution. To answer these questions, we need to look a bit deeper and understand the components of Cognitive Computing.

From a Solution Architecture perspective, we can simply think of Cognitive Computing in an evolutionary manner as an amalgamation of the state-of-the-art of proven technologies as Big Data Analytics, Machine Learning and Natural Language Processing. Figure 1 below shows a longer, yet not a compressive, list of the technologies and associated developments that differentiate Cognitive Computing from the previous generation of AI.

Among all the above innovations, one stands out as the key differentiator and that is the data. As James Kobielus stated Cognitive Computing is simply “AI that feeds on big data”. In addition, what has helped a lot is the over million-fold increase in computing power and much more in storage and data. Here lays the crucial difference: in the previous generation of AI we captured knowledge from human experts and coded them by hand now we can ‘discover’ knowledge (at least, a significant portion of it) from data and text.

Does this make sense to you? I would be very interested in your views on Cognitive Computing, its value and how you think it can be used to solve problems in your domain.