AI Safety Melbourne - Launch — LessWrong