Cognitive Technologies and Automated Analytics: Is There a Difference?

    January 27, 2016
    avatar

    By Dino Balafas

    Machine learning and AI have gradually moved from theoretical discussion to a real-life, everyday business application. But this increasingly ubiquitous technology goes by many names, and businesses must be careful to select performance over hype.

    Artificial intelligence has leapt from the pages of popular sci-fi novels and screenplays and into our everyday lives. Consider, for instance, the software I’m using to transcribe this paragraph as I dictate it out loud — it’s a simple enough task for a human, but one that requires some extraordinary computation for this technology. A natural language processor must categorize complex audio data as either useful or irrelevant, analyze it with probabilistic reasoning, and synthesize the result: in this case, the words you’re currently reading.

    As expected (minus some strange punctuation), this “cognitive technology” was spot on, as it is for millions of people every day. As this and other AI-enabled features like Siri attest, machines have become extremely accurate with pattern recognition and analysis.

    For identical reasons, cognitive technology has made a huge splash in business — it just topped the International Institute for Analytics (IIA) list of top 2016 analytics trends, and is expected to “subsume” automated analytics. A tweaking of the above example explains why: instead of audio information, computers assess a company’s enterprise data, plumbing it for patterns, and spit out readymade analyses, usually in the form of business decisions.

    All this is done at speeds and levels of accuracy no human could ever hope to match — it’s AI for business, in bright marquee letters.

    But hold on a second. Those functions — making data sets actionable and automating IT solutions — are squarely within the domain of automated analytics, especially the subsets of predictive and prescriptive analytics. Can cognitive technology really “subsume” an identical process?

    Is this just a clever marketing trick? A case of convergent product evolution? Is one of them AI and not the other? Do their results differ? What about their methods?

    The answer is: no, not really — but shifting conceptions on the role of analytics have made these questions more complicated than one might think.

    All Watched Over by Machines of Loving Grace

    Despite the popularity of AI as a topic, it can be difficult to pin down what exactly the term signifies. For decades we’ve imagined AI as an omnipotent, human-like computer consciousness that sways uncomfortably between benevolent savior and agent of nightmarish destruction.

    Obviously, this is not a product that a sane software consumer would want to purchase, let alone a savvy business. A more useful definition comes from Deloitte, which describes AI as, “computer systems able to perform tasks that normally require human intelligence.” Okay, reasonable: under this umbrella, both cognitive technologies and automated analytics more than qualify.

    But in fairness, while many different types of analytics exist, only a narrow set of the most sophisticated variety are really comparable to “cognitive technology.” But comparable how? While each of them is often pitched as the only bleeding edge competitor in the space, they have common origins.

    A Look Back and Under

    Automated analytics and cognitive technologies have their roots in the concept of machine learning, according to TechTarget, first conceptualized in the 1960s, but not practically effective until it became possible to process and store big data quickly and efficiently (hence the current resurgence).

    Rather than be governed by an extensive rules-based program (like most computers), these processors can identify, define, and analyze patterns on their own after being exposed to vast data troves — many of which could never be detected by human programmers. In truth, the quality of the result is largely determined by the sophistication of the algorithms guiding the process, but today, some algorithms have become incredibly sophisticated.

    Cognitive computing has sought to distinguish itself through deep learning via neural networks, a model borrowed loosely from the human brain. Very simply stated, deep learning allows computers to process many layers (10+) of complex data sets simultaneously. Show a deep learning computer 10,000 images of a cowboy hat and 3,000 without one, and it will form a more perfect understanding of a sombrero than you could ever hope to dream of — the same is true with IT trends of every kind.

    It’s impressive technology — but it’s not just limited to cognitive products.

    Coming Together

    In fact, as IIA co-founder Tom Davenport notes (via TechTarget), analytics professionals have long used deep learning, neural networks, and related technologies like logistic regression in their products. Put another way, there’s scant difference between the hardware undergirding these two tools; they’re the same. Still, buying one over the other will get you a different product — or rather, products.

    As he explains, the phrase “cognitive technology” has actually come to imply a package deal: automated predictive and prescriptive analytics, along with other tools that help organizations understand their IT data in a business context.

    For example, tools like business value dashboards can take those analytic results and summarize them in terms of KPIs that business leaders both understand and are deeply interested in. Or multiple types of analytics are housed under a comprehensive information management suite — many products masked as one.

    So, finally, is there a difference between cognitive tech and automated analytics? Not if we’re talking about analytic power and business application; but the answer might differ if we decide to define our terms differently. No matter the title, companies can only expect top-of-the-line performance if the applications they purchase are rock solid — and might we repeat, not every algorithm is created equal.

    But unlike humanity’s presumed fate under an AI overlord, the business benefits of machine learning are far from cloudy, measured in clear investment ROI and executive smiles (soon to be a business requirement). If only the terms we used to discuss it had such clarity.