Abeba Birhane has made it her life’s mission to expose the potential harms of Big Tech and AI, stemming from the data sets used to teach AI systems. Born and raised in Bahir Dar, Ethiopia, she moved to Ireland to pursue a Ph.D. in cognitive sciences at University College Dublin, and when she was surrounded by software developers, became intrigued by the data sets they were using. Upon inspection, she noticed the datasets had flaws. For example, the 80 Million Tiny Images dataset, used to teach machine learning systems how to recognize people and objects, had offensive labels, including racist slurs for images of Black people. Similarly, the ImageNet dataset contained pornographic content, with no explicit permission from those photographed. Birhane and her colleague, Vinay Prabhu, published their studies about these data sets, and the MIT team responsible eventually took it down and apologized.
This type of digital colonialism, in which powerful Big Tech companies export system philosophy and uses to the Global South without consulting those populations, mirrors the ugly reality of the hegemony of the certain members of the population, particularly in regard to race. Unfortunately, this bias has already resulted in real-world harms like Black patients being underserved in the medical arena, and a decreased likelihood of people of color obtaining mortgages. Birhane has even gotten the opportunity to present these issues to the Dalai Lama.
Big Tech firms have an enormous amount of power over the events in the real world, yet research is overwhelmingly male, white, and focused on problem solving rather than problem prevention. However, Birhane hopes to inspire individuals and organizations within the industry to take responsibility for recognizing when datasets are imbalanced and recognizing the potential harms of AI. She is determined to make sure Big Tech honoring those that it affects rather than simply trying to help them.
