Joseph Weizenbaum built ELIZA in 1966, then wrote “Computer Power and Human Reason” in 1976 arguing that some decisions should remain human. How do you see that early caution reflected — or missing — in today’s conversations about model accountability and societal impact?
Mandate audit trails and refusal policies for high-stakes uses; honors ‘some decisions should remain human,’ unlike today’s defaults…
Feels like the 1976 “some decisions should remain human” translates to a simple rule: if a model touches credit, hiring, or health, there must be a same-day human appeal with published monthly overturn rates. Building on @carly_smith82, make the appeal SLA binding and visible in the UI, so it’s not just policy but a real path to a person. Caveat: don’t bolt this onto low-stakes features where speed matters, but keep it mandatory for rights-impacting decisions.