Three of my beliefs about upcoming AGI — LessWrong