This month we're talking to Rohin Shah about alignment problems in AI, constraining AI behaviour, current AI vs future AI, recommendation algorithms and extremism, appropriate uses of AI, the fuzziness of fairness, and Rohin’s love of coordination problems.