For Part 1, we want you to explore the field and come up with your own project ideas, independently and without outside influence. Afterwards, there will be opportunities to collaborate, discuss ideas with other participants, and attend speaker events. We also know many of you are currently enrolled in the Cambridge AGISF program, so we hope to keep the March workload lighter and self-paced.
This document contains a list of papers and blogs related to the main subtopics in AI Safety. Choose 3-5 articles to read depending on what most interests you.
You do not have to choose your readings from this doc. If there’s another reading you’d prefer, please do that one!
For each reading, complete the Worksheets provided below.
The Worksheet includes writing down any questions you may have, such as: what does this concept mean? What would happen if we changed [x]? You may find it useful to reference the EA Forum post, Getting started independently in AI Safety.
Write down any project ideas that come to mind from the readings. We recommend framing ideas in the following format, from The Craft of Research: ‘I’m studying X, because I want to find out Y, so that I (and you) can better understand Z.’
Example: I’m studying the use of a discriminator in imitation learning, because I want to find out how to help humans produce demonstrations that the agent can imitate, in order to help my reader understand how we might use imitation learning to solve the reward engineering problem (from Richard Moehn)