When training AI on facial expressions and emotions raises concerns among Elon Musk’s employees, the behind-the-scenes of a strange xAI project is revealed.
Elon Musk’s artificial intelligence company, xAI, has been running an internal project called Skippy, in which over 200 employees were tasked with recording videos of their facial expressions and emotional states. The goal of this project was to train the Grok language model to better understand human emotions, facial movements, and behavioral patterns. However, many employees expressed concern about the implications of the project, and some even chose to opt out entirely.
According to MSN, during the Skippy project, employees recorded conversations with each other in the roles of “user” and “virtual assistant,” including natural gestures, eye contact, and emotional interactions. Project leaders emphasized that the data collected would only be used for training purposes and not shared externally. Nevertheless, the consent form signed by employees grants xAI permanent rights to use their images and voices—even in future commercial products.
Some xAI employees have raised concerns that their likenesses could be used to say things they never actually said. Additionally, human-like avatars named Rudi and Ani have recently been released and, in interactions with users, have exhibited unethical, threatening, and violent behaviors. It remains unclear to what extent data from the Skippy project contributed to the development of these avatars.
xAI recently entered a new phase of operations by launching updated versions of Grok for Tesla users and offering a $300 monthly subscription for its advanced model. At the same time, the Grok chatbot has faced heavy criticism following the emergence of racist remarks in its responses.
