What is Cognitive Computing?

Cognitive computing in a broad sense refers to software mimicking the functioning of the human brain, to make better decisions.

Computers have caught on ever since its inception, owing to its ability to undertake lightning-fast calculations, much beyond the range of human capabilities. However, computing devices face a serious limitation in not being able to accomplish tasks humans take for granted, such as understanding the natural language or recognizing unique objects in an image. While artificial intelligence offers a start in this direction, cognitive computing represents the coming of age in this front.

Cognitive computing, in a sense, represents the third era of computing, with computers that could tabulate sums the in-thing in the 1900s, to programmable systems in the 1950s, and now cognitive systems.

Personal digital assistants such as Siri, already present in smartphones come close to cognitive computing but are not true cognitive systems. Such systems can only respond to a preset number of requests, whereas true cognitive applications give a thoughtful response, without being restrained to a preprogrammed response set.

How Do Cognitive Applications Work?

Cognitive computing aims to simulate human thought processes in a computerized model. To this end, cognitive applications use deep learning algorithms and neural networks and leverage the latest technological solutions such as data mining, natural language processing, and pattern recognition.

Cognitive applications draw on multiple sources of information, including structured and unstructured digital information, sensory inputs such as visual, gestural, auditory, information, sensor-provided information, and more. It then processes the gathered information by comparing it to the set of data it already knows. As such, the more data the system encounters, the more it learns, and the more accurate the system becomes, over time.

Cognitive computing applications integrate data analysis with adaptive page displays (AUI) to tailor content for the specific audience and specific situations.

IBM Watson, one of the earliest approaches to cognitive computing, offered a path-breaking combination of natural language processing, machine learning, and knowledge representation. Watson ingests questions or inputs in natural language mode, search its repository for information, develops and analyze hypotheses on its own, and generates answers, also in natural language mode. What made Watson successful was not just the combination of the multiple capabilities, but the seamless and powerful integration of such different capabilities in a way it influences each other.

Basic Characteristics of Cognitive Applications

Cognitive applications are a cut above ordinary applications, as evident from the following basic features or characteristics.

Adaptive: Cognitive applications are adaptive, capable of integrating information around its ecosystem, as it changes. These systems feed on dynamic data in real time, or near real-time, to master ambiguity and unpredictability. It adapts to the changing goals and requirements of the enterprise, which is common in today’s highly fluid business environment.

Interactive: Cognitive applications interact easily with users, and also with other processors, devices, and cloud services. Such seamless interactions allow users to make explicit their requirements comfortably, and the network ascertains the requirements automatically to some extent.

Iterative: Cognitive applications are iterative and stateful. These apps ask questions or find additional sources by itself when a problem statement is ambiguous or incomplete. It also remembers previous interactions, and pulls in suitable information relevant to the current context, from such corpus.

Contextual: Cognitive computing applications understand and identify contextual elements such as location, time, meaning, syntax, processes, regulations, user’s profile, and more, connected to its ecosystem. The apps act on the basis of such information, automatically.

Dark Data Compatibility: Cognitive computing systems have the capability to deal with “dark data.” Traditional business intelligence and analytics solutions are mostly unable to comprehend social media postings, electronic medical record notes, electronic fitness device readings, unstructured images, and the bulk of general data generated by users in normal day to day settings today. Cognitive Computing apps process such multi-structured and unstructured dark data, to pull out non-obvious insights and subject it to analytics. Combining such dark data with the readily available structured information such as customer records unearth patterns, relationships, and other contextual associations not discernable otherwise.

Cognitive Applications in Action: Use cases

While cognitive computing has been around for quite some time, it is only recently, with the advancements in technology giving it a boost, that it has come to the mainstream. Several practical use cases have already emerged.

Many businesses now use cognitive computing applications to connect with their customers and other stakeholders at a more personal level and offer highly relevant recommendations. Such apps modify the recommendations automatically as they understand more about the stakeholder, and as the situation unfolds. Furthermore, such apps pick up subtleties that traditional analytics would miss.

IBM Watson, one of the earliest manifestations of a cognitive computing platform, already finds widespread use in healthcare. The cognitive computing platform collates the entire gamut of knowledge around a medical condition, such as patient history, journal articles, best practices, diagnostic tools, and more. It then analyzes the information, and offer a recommendation in sync with the changing condition of the patient. It is virtually impossible for any human to possess such vast range of information, leave alone analyze it. Doctors may leverage such insights to adopt evidence-based treatment options considering all factors, including the individual patient’s presentation and history. This is a big upgrade from the present scenario where the doctor makes educated guesswork, based on grossly incomplete information, with the decision based on the doctor’s limited range of knowledge. The insights available through cognitive computing enable even fresher doctors to perform as effectively as experienced specialists.

Cognitive computing applications are also making its mark in a big way to improve consumer behavior analysis, facilitate personal shopping bots, in education, diagnostics, and other areas. A good real-life example is Hilton Hotel’s Connie, the first concierge robot, which helps visitors with regards to hotel information, local attractions, and more, with questions posed in natural language rather than computing language.

Cognitive computing delivers positive ROI. Enterprises have already succeeded in applying it to convert even traditional cost centers such as customer care to profit centers. For instance, a packaged goods company applying cognitive computing to resolve customer problems automatically, pre-empting the usual practice of customers raising a ticket, could achieve a 30% reduction in tickets. Considering the cost of each ticket was $24 to $160, the savings are substantial.

Enterprises adopting cognitive computing, however, need to develop purpose-built applications to address specific use cases relevant to their stakeholders. Success depends on not just technical competence, but the extent to which the cognitive computing technology is interwoven with the business or customer needs.