At Data Summit 2019, Susan E. Feldman, president, Synthexis, and David Bayer, executive director, Cognitive Computing Consortium, offered an overview of key cognitive computing considerations in a highly interactive workshop titled “Cognitive Computing 101.”
According to Feldman, cognitive computing makes a new class of problems computable and addresses complex situations that are characterized by ambiguity and uncertainty. It handles human kinds of problems, and developers a human-like understanding of context and meaning. Cognitive computing systems redefine the nature of the relationship between people and their digital environment.
Cognitive computing is different from AI and it is important to understand how in order to know which to use when. Cognitive computing pillars that are shared with AI include the ability to learn and be adaptive, being probabilistic, and using big data from diverse sources. Characteristics that are specific to cognitive computing include being meaning-based, interactive, contextual, iterative and stateful, and highly integrated.
AI is meant to be autonomous. The machine is a substitute for the brain, whereas with cognitive computing the machine is an agent and partner, an information tool, explained Feldman.
Bayer highlighted a number of advantages offered by cognitive computing and why it is needed. In terms of scale, problems and data are complex and voluminous; it supports and augments human judgment with big data insights, and it leverages sources that are diverse and semantically rich. In terms of the probability of correctness, there is no single right answer, the context of the user and the data determines the relevance, it can offer ranked recommendations, and invites human-machine dialogue. In terms of discovery, it can also uncover unexpected relationships, trends, and patterns to spur innovation, possibilities and solutions, and it requires relationships among words, objects, and people.
Feldman and Bayer’s exploration of cognitive computing covered the key features, capabilities, and technologies that can be built into cognitive computing projects, such as text analytics, temporal information, voice recognition and analytics, image and visual analytics, machine learning/deep learning, classification and clustering, search, similarity matching, and others.
Feldman has been working on a framework for cognitive computing as well as a series of decision tools to help people understand the tradeoffs they have to make in deciding what kind of cognitive computing application they require. She presented an exercise to attendees to guide them in selecting the right cognitive computing application to fit their needs. The exercise offered an approach to evaluating advantages, possibilities, and compromises that may be faced in a project, and is a way to discuss the issues involved in the consideration of the scope of a project, she noted.
Many presenters are making their slide decks available on the Data Summit 2019 website at www.dbta.com/DataSummit/2019/Presentations.aspx.