The history of artificial intelligence has been dominated by a singular, somewhat narrow definition of intelligence: the capacity for logical reasoning, pattern recognition, and computational efficiency. From the early days of symbolic logic to the brute-force triumphs of Deep Blue in chess, and eventually to the generative prowess of Large Language Models (LLMs) like GPT-4, the trajectory has been one of cognitive scaling. We have built machines that can out-calculate, out-memorize, and now, out-write humans. Yet, for decades, these systems remained functionally autistic in their social capabilities, blind to the emotional subtext that constitutes the majority of human communication. They could process the text "I'm fine," but they could not detect the trembling voice or the averted gaze that indicated the speaker was anything but fine.
