Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Cognitive computing and big data analytics
Hurwitz J., Kaufman M., Bowles A., Wiley Publishing, Hoboken, NJ, 2015. 288 pp. Type: Book (978-1-118896-62-4)
Date Reviewed: Oct 5 2015

The claim that a system is “cognitive” can mean one of two very different things. For a half-century, the artificial intelligence (AI) research community has used the term to refer to approaches that mimic human mechanisms in reasoning. In this historic sense of the term, a cognitive system is based on some model of human reasoning (typically proposed by people called “cognitive scientists”) and contrasted with mechanisms that make no pretense to use psychologically plausible mechanisms. Examples of cognitive systems include coarse-grained symbolic reasoning systems such as Soar, as well as neural architectures inspired by biological models of brain function. Examples of alternative approaches include Bayesian inference and statistical machine learning. Those alternative mechanisms can produce excellent results and are often preferred because they are computationally more efficient, but ten years ago nobody would have called them “cognitive.” The inefficiencies of cognitive computing in this historical sense of the word make the title of this book irresistibly tantalizing: how could one apply such mechanisms to big data analysis?

Recently, IBM has begun using “cognitive” to describe its Watson system. In this context, “cognitive” no longer means reasoning the way that people do, but simply achieving results competitive with those that people achieve (as Watson has done impressively with its success on Jeopardy). Like DeepBlue before it, Watson’s internal processes have nothing to do with human reasoning, but are driven mainly by statistical analysis. Such statistical methods (historically called “weak methods” in AI) are not challenged by large amounts of data, but in fact require them to provide sufficient training instances; in this light, the title of the book is no longer a paradox, but a tautology.

This book has nothing to do with cognitive systems in the historic sense of the term. The closest it comes to cognition is its discussion of natural language processing (NLP) interfaces and the role of ontologies to organize knowledge bases, but the processing mechanisms it advocates are statistical. One of its longest chapters is a description of Watson, and the book might be described as a tour of Watson under the hood. The first two authors are principals in Hurwitz and Associates, and the company’s motto, “identifying the business value of technology,” aptly describes the technical level of the text. Complex diagrams (such as Figure 1-2) drop lots of names, but leave the reader to figure out what the geometric relation of the boxes to one another means. Technologies are described superficially and defined imprecisely (as in the case of “corpus,” or the distinction between ontologies and taxonomies, or the misleading example of “state variable” on p. 75). In spite of the authors’ recommendation on page xx of the introduction that readers should “gain more depth by exploring each topic in detail,” the book contains no bibliography or pointers to the research literature to aid readers in taking this advice.

The first seven chapters discuss component technologies, such as machine learning, natural language processing, ontologies, and cloud computing. Chapter 8 discusses the business implications of these technologies, chapter 9 describes Watson as a premier example of the kind of system the authors have in mind, and chapter 10 outlines the process of building such a system. Chapters 11 and 12 review applications in healthcare and smart cities, respectively. The last two chapters appear to have their titles reversed. Chapter 13, “Emerging Cognitive Computing Areas,” talks about upcoming application areas, while chapter 14, “Future Applications for Cognitive Computing,” really looks at emerging technologies that will extend the kinds of systems they have in mind.

This volume will be useful to business readers interested in a high-level overview of the technologies used in Watson and how those technologies might apply to their markets. Technical readers who want to learn either about cognitive systems in the classical sense of the term, or about Watson in particular, should look elsewhere.

More reviews about this item: Amazon

Reviewer:  H. Van Dyke Parunak Review #: CR143816 (1512-1024)
Bookmark and Share
  Reviewer Selected
Featured Reviewer
 
 
Statistical (I.5.1 ... )
 
Would you recommend this review?
yes
no
Other reviews under "Statistical": Date
A formulation and comparison of two linear feature selection techniques applicable to statistical classification
Young D., Odell P. Pattern Recognition 17(3): 331-337, 1984. Type: Article
Mar 1 1985
Remarks on some statistical properties of the minimum spanning forest
Dubes R., Hoffman R. Pattern Recognition 19(1): 49-53, 1986. Type: Article
Dec 1 1987
Statistical pattern recognition--early development and recent progress
Chen C. International Journal of Pattern Recognition and Artificial Intelligence 1(1): 43-51, 1987. Type: Article
May 1 1988
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy