Rooting out racism in AI systems -- there's no time to lose

How will AI strategy in the enterprise be changed by the widespread attention to systemic racism? Like a lot of complicated topics, the discussion of racism in AI systems tends to be filtered through events that make headline news -- the Microsoft chatbot that Twitter users turned into a racist, the Google algorithm that labeled images of Black people as gorillas, the photo-enhancing algorithm that changed a grainy headshot of former President Barack Obama into a white man's face. Less sensational but even more alarming are the exposés on race-biased algorithms that influence life-altering decisions on who should get loans and medical care or be arrested. Stories like these call attention to serious problems with society's application of artificial intelligence, but to understand racism in AI -- and form a business strategy for dealing with it -- enterprise leaders must get beneath the surface of the news and beyond the algorithm. "I think that racism and bias are rampant in AI and data science from inception," said Desmond Upton Patton, associate professor of sociology at Columbia University. "It starts with how we conceive a problem [for AI to solve]. The people involved in defining the problem approach it from a biased lens. It also reaches down into how we categorize the data, and how the AI tools are created. What is missing is racial inclusivity into who gets to develop AI tools."