Machine learning, just objective computation?

A friend recently told me a story, her 6 year old nephew asked Alexa (Amazon Echo personal assistant),  “What’s my grandfather’s name?” My friend, who happened to be in the same room, answered his question. Few days later, her nephew asked the same question again and Alexa answered correctly.

Machine learning a discipline combining science, statistics and computer coding that gives computers the ability to learn, diagnose, predict, and plans without being explicitly programmed. Such capacity is driven by algorithms that teach themselves to grow when exposed to new data. Unlike artificial intelligence that develops the cognitive capabilities to rival human intellect, machine learning is more about optimizing certain problem-solving processes.

Machine learning produces models that can analyze big, complex data and provide quick and accurate results. Such precise models allow organizations to identify profitable opportunities – or avoiding unknown risks. For instance, websites recommending items you might like based on previous purchases are using machine learning to analyze your buying history – and promote other items you'd be interested in. This ability is the future of retail. We’ve all heard of the famous story when target sent a teenager ads for baby products while her father didn’t even know is own daughter was pregnant. This is possible due to models analyzing your purchase patterns and predicting future behavior.

While this is all very impressive, we need to keep in mind where these data come from: your digital crumbs, even the ones that you have not disclosed. They can infer your sexual orientation, your personality traits, your political leanings. They have predictive power with high levels of accuracy. Another problem is the gathered data are human imprints hence reflecting our biases, which makes machine learning not an objective, neutral computation. For instance on Google, women are less likely than men to be shown job ads for high-paying jobs. Searching for African-American names is more likely to bring up ads suggesting criminal history, even when there is no such evidence. Such biases amplified and showed back at us can only deepen our stereotypes if we don’t digest these with a grain of salt. It can also have life-altering consequences.