It's nice to see a discussion of AI risk that addresses concrete scenarios. A lot of the forecasted doom (in other reports) resorts to handwavy arguments, but r

It's nice to see a discussion of AI risk that addresses concrete scenarios. A lot of the forecasted doom (in other reports) resorts to handwavy arguments, but rarely goes into specifics. The examples they've given[0] seem plausible enough to me (except the persuasive ads one). [0]: Persuasive ads, vulnerability discovery and exploitation, hacking robots (this one is only tangentially AI related), and AI-augmented surveillance.