Forecasts for AI-driven job displacement range from "gradual over decades" to "most cognitive work within years." The methodologies differ wildly and the incentives to hype or downplay are strong. Which forecasts do you actually trust, and what evidence would change your mind?