Google Releases Open AI Models With Emotion Detection, Alarming Experts
Google launches the PaliGemma 2 family of open AI models that claim to identify emotions in images. Critics warn the technology relies on flawed science and risks spreading harmful biases.
Google introduces the PaliGemma 2 family of open AI models, which boast the controversial ability to identify emotions in photos. Built on the Gemma 2 series, these models analyze images to generate detailed captions that describe actions, emotions, and the overall narrative of a scene. While the emotion recognition feature requires fine-tuning to function, its mere availability in an open-source format raises significant concerns among technology ethicists.
Experts express deep skepticism about the foundational science behind emotion-detecting AI. Sandra Wachter, a data ethics professor at the Oxford Internet Institute, compares the technology to relying on a Magic 8 Ball, emphasizing that humans cannot simply "read" emotions from facial expressions. This skepticism stems from the fact that most emotion detectors rely on outdated psychological theories that ignore how cultural and individual differences drastically alter how people express their feelings.
Consequently, critics warn that deploying such systems leads to unreliable and highly biased outcomes. As past research demonstrates, face-analyzing models routinely develop unintended preferences for certain expressions and fail to account for the complex, nuanced nature of human emotion. Because these tools can never fully "solve" emotion detection, experts worry that making them openly available invites misuse by marketers, spy agencies, and other entities seeking to exploit flawed data.