UCLA Study: AI Lacks Human-Like 'Internal Embodiment'

Researchers at UCLA Health published a study in the journal Neuron regarding artificial intelligence safety ^[1]. The team argues that current multimodal AI systems lack internal bodily awareness, such as fatigue or uncertainty ^[1][2]. These "internal embodiments" are critical for aligning AI behavior with human behavior and safety needs ^[1]. While AI can accurately describe human feelings, it does not actually experience them ^[2]. The study suggests that this gap prevents current systems from fully understanding human safety requirements ^[1][3]. Experts propose that future AI might require simulated internal sensors to better mimic human physical states and improve reliability ^[3].