Category: General AI safety monthly meeting - interpretability of Adam Shai's approach