We are a group mostly consisting of students and beginning researchers in artificial intelligence. We are interested in how to build safe artificial intelligence and regularly have reading meetups with in-depth discussions about this. Papers and blog-posts we discussed so far were about the following:
- Expert predictions about the arrival of artificial general intelligence.
- The "Concrete Problems in AI Safety" paper. In particular:
- Negative side-effects in reinforcement learning and solution attempts through "Relative Reachability".
- The "Agent Foundations" framework of the Machine Intelligence Research Institute.
Join our in-depth reading meetups for people interested in working on aligning algorithms with human values.