LLMs definitely present dangers. At the same time, I think there might be more present, tangible risks such as indirect model biases in decision making and enfeeblement when models are used as substitutes to everyday life functions such as AI emotional companions and over reliance to copilots.