Tegmark's 2017 piece on AGI and existential risk — still very relevant given recent developments in AI.