Existing comment:
Machine Vision
What are artificial intelligence systems actually seeing when they see the world?
-- Trevor Paglen, 2017
The advent of digital pictures marked the transition from human-seeable to machine-readable images. Electronic images are merely data to be processed by computer programs. For example, human eyes cannot see a .jpg file without a web browser or photography editing tool to "translate" the data. Because we rely on such tools for so many aspects of life, surveillance -- whether by government, corporate, or other interests -- has moved to our most intimate spaces -- desktop computers, tablets, mobile phones. In addition, artificial intelligence (AI) now routinely categorizes and analyzes machine-generated images. The development of automated seeing systems -- such as those in facial recognition and self-driving cars -- concerns Paglen. What is radical about this moment, he wrote, is that most images are now made by machines for other machines. Increasingly AIs carry out operations without any human "seer" or human intervention. Such systems challenge our understanding of what images are and what they do. With his Machine Vision works, Paglen raises questions of humanity. What does it mean that "seeing" no longer requires a human "seer"? If images are a defining characteristic of human history and culture, will machine images redefine humanity? Or will images become its epitaph? |