Research Scientist

• Advance mechanistic interpretability by developing data structure models and synthetic datasets to benchmark AI interpretability tools.
• Develop tractable, scale-aware data structure models grounded in physics theory.
• Quantify how features learned by AI systems relate to underlying data structures.
• Conduct interdisciplinary research projects bridging physics and AI interpretability.

This role directly contributes to understanding and improving the transparency of AI systems, which is crucial for ensuring AI safety and alignment with human values.

• Strong background in physics, mathematics, or related quantitative field.
• Experience with machine learning and AI interpretability research.
• Proficiency in programming (e.g., Python) and data analysis.
• Ability to conduct interdisciplinary research.

Compensation and benefits are competitive and commensurate with experience. Specific details not provided.

Principles of Intelligence is an organization focused on advancing the science of AI interpretability and safety through interdisciplinary research.