This research continues many of the directions our team worked on prior to Anthropic, including: GPT-3, Circuit-Based Interpretability, Multimodal Neurons, Scaling Laws, AI & Compute, Concrete Problems in AI Safety, and Learning from Human Preferences. Partner with FP&A, Accounting, Tax and business teams to understand forecast movements, challenge assumptions and improve accuracy, while strengthening processes, reporting and controls so the framework scales.