Why UX should guide AI

If we need to learn one thing about the numerous AI applications around us today, it is that they are examples of "artificial specific intelligence." In other words, they rely on algorithms that are great at very particular tasks, such as selecting a movie based on our watching history or keeping our car in the proper lane on the highway. Because it is so highly specialized, AI greatly outperforms human intelligence in those narrowly defined tasks. Take it from a person who recently spent 50 minutes picking a movie that itself lasted 77 minutes. However, AI's effectiveness at specialized jobs comes at the price of severe context blindness and a general inability to develop meaningful feedback loops: The typical algorithm does not and cannot consider the wider implications of the decisions it makes and hardly affords us users any control over its inner workings.