Unmasking AI: My Mission to Protect What Is Human in a World of Machines

Part memoir, part polemic, this trenchant debut from computer scientist Buolamwini chronicles her career studying encoded bias in artificial intelligence. She explains that her interest in the failures of AI started when, as a master’s student at MIT in 2015, she struggled to get a facial recognition program to track her face. Putting on a white theater mask did the trick, revealing that the program only worked for light-skinned faces and alerting Buolamwini to how human prejudices get baked into tech. Exploring the ethical quandaries posed by AI, she shares research that found facial recognition software’s spotty record detecting dark-skinned individuals’ faces stemmed from training the software with image datasets mostly comprising light-skinned people. Attempts to fix this bias have led to further injury, according to Buolamwini, who describes a China-based startup’s initiative to collect photos of Zimbabwean individuals’ faces for AI datasets as a form of “data colonialism” that exploits Zimbabwean people “to build the wealth of foreign companies.” Buolamwini proves that she’s among the sharpest critics of AI, and her list of principles for achieving “algorithmic justice,” which includes the stipulation that “people have a voice” in shaping the algorithms that influence their lives, charts a path forward. Urgent and incisive, this is a vital examination of AI’s pitfalls. [em](Oct.) [/em]