[Sydney] Paper Reading with Anthropic Research

Before "alignment" was a LinkedIn buzzword, a small team at Anthropic was writing the papers that defined it. Zac Hatfield-Dodds was in the room for most of them. We're sitting down with Zac and Relevance AI to go through 'Toy Models of Superposition.' Less lecture, more book club. For researchers, engineers, and founders working on or near safety :)Please read the paper before joining us! https://transformer-circuits.pub/2022/toy_model/index.html Agenda: 1:30 - 2:00pm | Registration 2:00 - 2:30pm | Paper Presentation from Zac 2:30 - 4:00pm | Q&A Discussion