Hollywood has done a good job of convincing us that very soon robots are going to take our jobs and run our world. But in real life, artificial intelligence poses a more pressing danger: bias.
My fellow AI technologists need to fix this problem urgently to prevent the algorithms and machines we create from replicating the human mistakes and casual prejudices the world has come to accept as normal.
Taking bias out of AI starts with the nomenclature. Giving AI-driven technologies gendered names, for example Amazon’s Alexa and IBM’s Watson, perpetuates long-held stereotypes. Gendered names can affect how users interact with technology. In these early days of voice-driven AI, studies have found that users are more likely to ask Amazon’s Alexa to play them a song, but IBM’s Watson to pull the data required to solve a difficult business problem.