Abstract: This presentation explores AI through
research on YouTube’s video recommendation system. I draw on empirical findings
from an ongoing study of video meme circulation on YouTube in order to expand
the concept of AI to include the many technical systems powered by machine
learning that individuals interact with on a daily basis. The study involves
tracing the flow of YouTube’s recommendation system in order to understand how
it treats video memes—a relatively recent type of content based on bottom-up
appropriation, remix, and re-circulation. How YouTube’s recommendation system
understands video memes can teach us about how machine learning systems
interact with cultural content that does not easily fit into pre-existing
genres and categories. As meme culture continues to expand—as evident in the
overwhelming popularity of TikTok, for instance—researchers of AI should pay
closer attention to the influence of the many recommendation systems embedded
in the infrastructures of today’s most dominant digital communication and media
technologies.