Can robots truly read the room? Cornell researchers are delving into this intriguing question, exploring the potential for artificial intelligence to bestow robots with social intelligence. The study, led by Maria Teresa Parreira, focuses on the ability of vision language models (VLMs) to predict the outcomes of tense scenarios in short videos. The research reveals a fascinating insight into the limitations of current AI technology in interpreting human facial expressions and using that information to anticipate outcomes. The findings highlight a significant gap in anticipatory social intelligence among current VLMs, emphasizing the importance of developing robots alongside humans for successful integration into society. This article delves into the complexities of human-robot interaction and the challenges that lie ahead in achieving true social intelligence in machines.
