Machine learning has captured the human imagination, but the notion that machines will somehow replace humans is an idea best left to science fiction.
The Importance of the Algorithm
Algorithms have played a key role in ad tech since the very beginning of the industry. However, the simple fact is that we don’t differentiate between companies based on their algorithms is unsettling at best and dangerous at worst. For one, the algorithm is proprietary, so there are legal limits on the ability of marketers to compare machine learning technology. But the most salient limit is a practical one. The math is simply too complex for 99 percent of marketers to grasp the nuts and bolts of the algorithm in any tangible or meaningful way. Is that a bad thing? Not really, no. Machine learning plays an important role in marketing, but it isn’t synonymous with marketing. Marketers vet their vendors based on input and output. They want to know what types of data go into building an audience and how effective the results of that data collection and analysis are to achieving their larger business goals. Marketers ask these kinds of questions from algorithmic solutions because marketing hinges on strategic and creative thinking that addresses real life challenges and opportunities. A machine can optimize a media buy, but it can’t articulate the emotional connection the brand is striving to make with consumers.
Machine Learning’s Ideal Purpose
Machine learning offers the ability to process large data sets and create predictive modeling around that information. This is incredibly valuable, but only within the context of a larger human endeavor. Medical researchers recently used machine learning to create the most detailed map of the human brain to date—a development that many believe will revolutionize medicine. But with that breakthrough, the question becomes “what will doctors do with this treasure trove of information?” In assessing the impact of machine learning and big data on the healthcare field, some in medical professionals argue that we’re moving away from group-based medical solutions towards a more individualized standard of care. If that’s true, machine learning not only arms doctors with greater insight, it also makes the role of those doctors that much more relevant because additional information will lead to a wider range of options and decisions.
Humans Are Always Necessary
Among IT executives, it’s well-understood that machine learning provides enormous value, especially around areas of predictive analytics, recommendation systems, and cluster analysis and segmentation. But among workers directly impacted by those advances, there is a persistent fear that machine learning will replace humans. History proves that fear is overblown. Yes, automation on the assembly line meant fewer workers were necessary to perform the task at hand, but didn’t that automation also free up many workers to perform other functions, or join entirely new industries? Several years ago, programmatic trading looked like it would kill the media buyer. Of course, in hindsight, we know that wasn’t the case. Programmatic trading reduced the number of salespeople making cold calls as well as the number of analysts necessary to calculate media yield, but those activities were at the margins of the advertising business. The core of the advertising business is to draw out emotional connections between brands and consumers. A media buyer intuitively understands what an algorithm can never grasp—that a Ford means something very different to an audience than a Cadillac. Programmatic tools do a lot of the heavy lifting to help the media buyer find and price an automotive audience and identify segments that are more likely to be interested in one type of car over the other, but ultimately every advertising campaign comes down to a human question—does it resonate? The human mind is incredibly well-suited to grapple with subjective questions like emotional resonance. We are always refining our critical thinking skills because we have thought and decision-making processes that can be followed, evaluated and scored. Put simply, human intelligence isn’t just more comprehensible than machine learning, it’s also far more adaptable, which makes us ideally suited for a dynamic world. Machines can and should help us uncover the trends that shape our world, but if we want to understand a human endeavor, we’ll always need to ask a person.
