{"component":{"type":17,"accent_color":4095858,"components":[{"type":12,"items":[{"media":{"url":"https://annarborai.org/assets/social/resources-paths-three-questions-about-ai-safety-0283dda48d649395.webp"},"description":"“Three questions about AI safety” over a watercolor of the Matthaei Conservatory Temperate House."}]},{"type":10,"content":"## [Three questions about AI safety: a reading path](https://annarborai.org/resources/paths/three-questions-about-ai-safety/)\nThree readings on AI systems missing their designers’ intent: DeepMind on specification gaming, Concrete Problems in AI Safety, and Anthropic’s circuit tracing."},{"type":1,"components":[{"type":2,"style":5,"label":"Join the A4 Discord","url":"https://discord.gg/bkHxEk4nHH"},{"type":2,"style":5,"label":"Take part","url":"https://annarborai.org/join/"}]}]}}
