When A.I. is More Artificial than Intelligent
How my industry’s reliance on buzzwords has done itself and our customers a disservice.
I am constantly energized by the lightning-fast pace at which the predictive analytics industry is evolving. I am proud to be a part of a company that is making a difference for our users. However, my job is so much more difficult than it needs to be because of the wounds that the predictive analytics industry continues to inflict upon itself in the form of marketing messages that only serve to confuse audiences thirsty for knowledge.
Good companies in the predictive analytics space are all fundamentally founded on the premise of helping business leaders and/or consumers efficiently make better decisions. When successful, these better decisions will result in finding previously unseen value. By reducing the guesswork involved in decision making, the value for the end user should be savings of cash or time, waste reduction or revenue increases. All of these are the intended consequences for consumers of certain predictive analytics tools. That’s a pretty simple premise.
And yet as an industry, we do everything to complicate the premise by refusing to engage in a basic dialogue of how our tools work. Instead, we rely on buzzwords to describe our products. Sometimes these buzzwords have actual definitions (AI, Proprietary Machine Learning Algorithms, et al). Other times we use or clichés (black box, secret sauce, et al).
More often than not, use of these buzzwords seems intended to avoid a reasonable discussion about what happens to a dataset to arrive at the insights generated by the predictive tools. The message is basically, “trust us, we’re smart.”
As money continues to pour into companies throughout the nascent analytics industry from investors and consumers alike, there is not a natural impetus to change this messaging. Yet.
“Trust us, we’re smart” won’t be a message that sells forever. Even analytics companies that are successful today with good products will be burned if our markets develop a strong enough distaste for our industry. That distaste will be sharpened by buyers of ineffective products when they learn that the sauce isn’t so secret, the black box is empty, and that the A.I. is more artificial than intelligent.
Some of the analytics companies that continue to rely on the buzzwords actually have products that are powered by truly ingenious mathematical approaches. Other companies use the buzzwords to avoid discussions about mathematical engines that a decent high school math student could produce.
Without a high level description of how the “black boxes” work, the buzzwords subtly connote magic. People love the entertainment value of magic acts because they appreciate that they can be tricked. However, when people face a decision of any financial consequence and they instinctively feel that magic is involved in a decision making tool, they quickly discount – or totally ignore – the output. The mystery that my industry has shrouded itself in by its continued overreliance on these terms has caused suspicion. That suspicion has led to underutilization of analytics tools by the people and the companies that need the most help.
Those of us who believe in the decision making power that our industry can bring – producers and consumers of analytics alike – are responsible for changing this environment.
Good analytical products are built on a foundation of solid mathematics, but we insult the intellect of our purchasing populations when we over-rely on buzzwords and too-good-to-be-true ROI claims. Sellers of these products need to ensure that their users understand how the products work. This doesn’t mean that I encourage anyone to publish their algorithms for the world to see, but we can all discuss the inner workings of our mathematical engines in a way that is accessible to broad audiences. Allowing users to understand the basic inputs that drive our models is a step in the right direction. Deeper discussions of how changes to these inputs impact the outputs will bring us a lot closer to credibility.
Until competition or a tightening market compels changes in analytics providers’ messaging, it will ultimately require the buyers of these products to demand answers to basic questions. “How does it work” seems like a good place to start.
Jim Dries is the CEO of piLYTIX, a provider of predictive analytics solutions for unique sales organizations.