Antonio Damasio — The Feeling-Machines Question

The most surprising thing Antonio Damasio ever did about artificial intelligence was not to dismiss it but to specify it. In 2019, with neuroscientist Kingson Man, he co-authored a paper in Nature Machine Intelligence asking, in earnest, what it would take to build a machine that genuinely cares about what it does. The paper argues that the trouble with conventional AI is structural: it is built to pursue goals without ever facing the question of whose goals or why, optimizing objectives handed to it from outside. A truly intelligent agent, Damasio and Man proposed, ought instead to hold its own meta-goal of self-preservation, the way living things do, with survival riding on continuous regulation of its own internal states under genuine risk. A machine built that way—one that could be damaged or destroyed and was organized around preventing it, with