The Danger of AI is Weirder Than You Think

Author: Janelle Shane

Publisher: TED

Publication Year: 2019

Summary: The following TEDx Talk explores the nature of relationships between humans and artificial intelligence (AI). Shane proposes that AI was a tool that humans (particularly data scientists) need to understand how to wield well. This relies on people’s ability to communicate with one another and then appropriately transfer that into commands that an AI would be able to follow. As they point out AI has no capabilities to understand what kind of data it is being given to them nor the implications of the routes the algorithm takes. Oftentimes there is not only logic behind the solutions that an AI developed: AI will do exactly what we tell it to. As a result, data scientists need to create very clear and cognizant limits when they decide to implement AI. This does not mean that optimization or another metric should not take precedence when creating a model. It does however mean that these considerations and questions need to be a part of the development process.