Super Crunchers: Why Data-Driven Thinking is the New Smart - Perfect for Business Analytics, Decision Making & Self-Improvement

This short book, which can be read in a few hours, could be considered an apology or even a manifesto for mathematical and statistical modeling. Even those readers, such as this reviewer, who have been involved in "supercrunching" for many years will find some interesting anecdotes in this book that illustrate its power and limitations. But even more importantly, it discusses the reactions of many (and typically highly insecure) individuals against the practice of mathematical modeling, some of these bordering on the absurd but with most content with ridiculing its practice. There is no question that the "supercrunching" that the author describes will continue to have greater influence in the manner in which it is currently practiced, but it is also true that much more sophisticated approaches to modeling will arise, some of these using machine intelligence and highly advanced mathematics. Indeed, in the past decade the use of artificial intelligence has exploded in areas such as finance, network engineering, bioinformatics, and Internet security. The author has just scratched the surface of the vast number of tasks that are now being done by machine, sometimes without any intervention or supervision by humans.As the author details in the book, many businesses and public institutions have jumped on the bandwagon of supercrunching, and their successes in doing so he documents well, with references given for readers who want more of the details. But many businesses that could profit greatly from this approach have refrained from its use, because of skepticism or distrust of quantitative reasoning. To paraphrase the author, they want to stay with the horse-and-buggy, while others are getting around in locomotives. There is still a great reliance on "experts" whose track record is weak and when compared with statistical modeling falls very short. It is an open question whether these businesses will find themselves in bankruptcy court because of their rejection of statistical modeling, but they are going to have to face stiff competition from the businesses that do.For those involved in the supercrunching that the author describes, it is not surprising to hear that many important business decisions are being made based on the results of statistical modeling. Humans are to a large degree still "in the loop", but the author describes instances where the machines "are actually in charge." The author still wants both human and machine to be in a mutual symbiosis, with considerably more weight given to machine predictions. And along these same lines, he brings up the canonical question as to what place humans are to have as more responsibility for decision-making is given to the machines. Many may find their social and employment status shrink, becoming white elephants (or "potted plants" to use the author's terminology) in the process. The author tries to alleviate these anxieties by pointing to the need for humans to still do the groundwork that enables supercrunching to take place. Humans must still "hypothesize" he asserts, in that they must still make the decisions as to what variables are going to be used when the machines actually perform the statistical analysis.Certainly if one remains within the statistical modeling paradigm, as the author does throughout the book, there will still be need for humans to "hypothesize." But if one chooses to go beyond this paradigm, the landscape changes considerably. The author gives a brief glimpse into how this is done in his discussion on neural networks. But he leaves out any discussion of the research in automated scientific and mathematical discovery that has taken place in the last two decades, some of this research showing remarkable progress. If this trend continues, and there is every reason to think that it will, then the machines will be able to hypothesize, theorize, and analyze in a manner that is similar to humans but may be vastly superior. In addition to these developments, significant progress has been made in artificial intelligence that allow machines to use defeasible, abductive, and inductive reasoning patterns in order to operate in domains unheard of just a few years ago. These domains include automated legal reasoning, computational creativity, rumor detection and propagation, virus recognition in data networks, musical composition, and automated mortgage underwriting. With these and their supercrunching abilities, they will certainly instill both admiration and fear, and using them will require extreme confidence to a degree that goes far beyond what the author describes in this book. It is disquieting to some that the machines will have this degree of intelligence and autonomy, but to others it is a source of pure exhilaration, and proof again that this is the best time ever to be alive.