Grant: Grant to "support a project to develop interactive AI algorithms for alignment that can uncover the causal features in human reward systems, and thereby help AI systems learn underlying human values that generalize to new situations." (FTX Future Fund → University of California, Berkeley)