Definitely in some ways, haha. I’m feeling pretty smug about some of my less-obvious predictions, e.g. the “routing around the censorship thing to find out what models really think and feel” thing is currently happening a bunch with ChatGPT. And Diplomacy has turned out to be easier than I expected, though to be clear I was imagining humans losing to the bot even after knowing they were up against a bot, which is NOT what happened with Cicero, Cicero would have lost under those conditions.
Definitely in some ways, haha. I’m feeling pretty smug about some of my less-obvious predictions, e.g. the “routing around the censorship thing to find out what models really think and feel” thing is currently happening a bunch with ChatGPT. And Diplomacy has turned out to be easier than I expected, though to be clear I was imagining humans losing to the bot even after knowing they were up against a bot, which is NOT what happened with Cicero, Cicero would have lost under those conditions.
What else did you have in mind?