In one of a series of essays on the impact of AI on the insights profession, Martin Rückert describes how machines can be trained to efficiently learn your business by building an understanding of marketing into the underlying knowledge graph.
If I showed you a picture of a cat you had never seen before, you would quickly be able to identify it as a cat. If I showed a 4-year-old a picture of cats and tigers, they would be able to distinguish between the feline categories. Cognitive computers, however, find the cat exercise immensely challenging. Unlike humans, who have evolved to quickly comprehend categories like “cat” and “tiger” from only a few examples, machines struggle to learn from a small number of training images.
Most machine learning algorithms need to ingest huge amounts of data about isolated concepts and then connect the dots. You would have to show the algorithm hundreds of thousands of examples of a cat for it to identify one it had never seen before. Through a process of intense computation, it would first learn to identify the unique features of cats on a more abstract level, adapt the decision criteria over several learning iterations, and then store those weighted rules. Despite the extensive cat identification training, if you showed it a picture of a tiger, it would likely classify the cat as a tiger as well.
Unlike computers, which need to learn new concepts from scratch, humans can base their object categorization identification on past learnings. A 4-year-old may have already identified a dog before seeing a cat, and can abstract that learning to a meta-level: “How is (category) cat different from (category) dog?” That is a much easier task, and over time, the speed of learning new concepts accelerates. The more abstractions we learn, the quicker we can spot the new features in a previously unseen concept — it’s estimated that by the time a child is 6 years old, they have already learned almost all of the 10-30 thousand object categories in the world. The human ability to identify categories is what enables a marketer to learn the difference between the concepts of product “problems” and “benefits.”
At Market Logic, we help our systems learn to categorize data from a small number of training examples by using prior knowledge about object categories, using a data science method called “one-shot learning.” Additionally, we train the machine as a managed service so that understanding of marketing is built right into the knowledge graph we call the Market Logic. It’s a key reason why Harvard Business Review says our insights engines deliver real value for clients like Unilever from day one.