Abstract: The AIs of Facebook, YouTube, and Tweeter control a large proportion of the information we see every day. Depending on small parameters, they could recommend radically different content. Do they favor a certain type of messages? By collecting open data on YouTube with the site algotransparency.org, we show that the goals of the AI and the user are not always aligned.
Bio: Guillaume Chaslot focussed his PhD on Monte-Carlo methods for Computer Go. After working at Microsoft, YouTube and Google, he created the website AlgoTransparency.org to to analyze the impact of YouTube's AI on major societal issues. He is now working in collaboration with the EU Disinfo Lab and the Center for Humane Technology.