I’ve been impressed by Yud in some podcast interviews, but they were always longer ones in which he had a lot of space to walk his interlocutor through their mental model and cover up any inferential distance with tailored analogies and information. In this case he’s actually stronger in many parts than in writing: a lot of people found the “Sable” story one of the weaker parts of the book, but when asking interviewers to roleplay the rogue AI you can really hear the gears turning in their heads. Some rhetorical points in his strong interviews are a lot like the text, where it’s emphasized over and over again just how few safeguards that people assumed would be in place are in fact in place.
Klein has always been one of the mainstream pundits most sympathetic to X-risk concerns, and I feel like he was trying his best to give Yudkowsky a chance to make his pitch, but the format—shorter and more decontextualized—produced way too much inferential distance for so many of the answers.
I’ve been impressed by Yud in some podcast interviews, but they were always longer ones in which he had a lot of space to walk his interlocutor through their mental model and cover up any inferential distance with tailored analogies and information. In this case he’s actually stronger in many parts than in writing: a lot of people found the “Sable” story one of the weaker parts of the book, but when asking interviewers to roleplay the rogue AI you can really hear the gears turning in their heads. Some rhetorical points in his strong interviews are a lot like the text, where it’s emphasized over and over again just how few safeguards that people assumed would be in place are in fact in place.
Klein has always been one of the mainstream pundits most sympathetic to X-risk concerns, and I feel like he was trying his best to give Yudkowsky a chance to make his pitch, but the format—shorter and more decontextualized—produced way too much inferential distance for so many of the answers.