Rating is available when the video has been rented.
This feature is not available right now. Please try again later.
Published on Apr 16, 2020
Humans of AI is about the people who make modern artificial intelligence (AI) possible: the commonly un-recognized, involuntary contributors of datasets that teach computers to see, listen, or talk. This talk tells the story of the creation of the COCO (Common Objets in Context) image dataset. It is a story about people. About love, live, and death. About curiosities such as ten male researchers voting amongst themselves on what is "common" and then consulting actual 4-year-olds to verify that they were doing it right. During the talk I present three interventions inspired by my investigation, not least as attempts to correct some of the issues I discovered. The data by which machine learning algorithms learn to make predictions is hardly ever shown, let alone credited. By doing both, Humans of AI exposes the myth of magically intelligent machines, instead applauding the photographers that made this technical achievement possible. In fact, when showing the actual training pictures, credit is not only due but mandatory.