Bio

Dr. Timnit Gebru is DAIR’s founder and executive director. Prior to that she was fired by Google in December 2020 for raising issues of discrimination in the workplace, where she was serving as co-lead of the Ethical AI research team. Timnit also co-founded Black in AI, a nonprofit that works to increase the presence, inclusion, visibility and health of Black people in the field of AI, and is on the board of AddisCoder, a nonprofit dedicated to teaching algorithms and computer programming to Ethiopian and Jamaican highschool students. She has received a number of accolades including being named one of Nature’s Ten people who helped shape science and one of TIME 100’s most influential people. She is currently writing The View from Somewhere, a memoir + manifesto arguing for a technological future that serves our communities instead of one that is used for surveillance, warfare, and the centralization of power by Silicon Valley.

Talk Abstract

In the last few years, the quest to build so-called Artificial General Intelligence (AGI), an undefined system which seemingly can do any task under any circumstance, has captured the public's imagination. Those whose mission has been to build this system, like the leaders of OpenAI, Anthropic, Deepmind and others, discuss the utopia that will imminently come from building AGI, or the apocalypse that will be caused by it rendering humanity extinct. In this talk, I discuss the history of the AGI movement, and its link to the 20th century eugenics movement, with those who "christened" the term AGI having the goal of replacing humans with a superior race they call "transhuman AGI." I outline the harms the quest to build AGI has caused, including labor exploitation, centralization of power and the safety issues associated with building an unscoped system. I close by giving examples of various movements to resist this trend, including artists' fight to preserve their humanity and dignity with the hashtag #TeamHuman. I urge the machine learning community to focus on small, constrained, task specific models, and present some of our work from DAIR showing how this approach outperforms the one size fits all trend to building machine learning based models.