Response to Katja Grace's AI x-risk counterarguments — LessWrong