Anthropic opens two research fellowships in AI, science, policy and law

Anthropic's new fellowships aim to integrate external expertise to test and improve its Claude AI models. This initiative suggests the company sees value in diverse perspectives, particularly from fields like philosophy and public policy alongside traditional STEM, to identify subtle weaknesses and enhance AI performance on complex problems.

The programs distinguish between a STEM Fellowship, focused on scientific and engineering challenges, and a Conceptual Reasoning Fellowship, which targets areas requiring specialist knowledge and judgment. STEM fellows will develop datasets and simulations for scientific reasoning, while conceptual reasoning fellows will create benchmarks for assessing model performance in areas like ethical or legal interpretation, where answers are often less definitive.

Anthropic is seeking highly qualified researchers who can contribute immediately, as indicated by the requirement for hands-on experience with Claude and Claude Code, and the absence of visa sponsorship. This approach suggests a focus on quickly integrating experienced individuals, likely with existing US work authorization, to achieve specific research goals within the fellowship's four-month duration.

The compensation range of $100,000 to $167,000 for the four-month term positions these fellowships as competitive opportunities for experienced professionals. It reflects the specialized knowledge and research background Anthropic expects, emphasizing a collaborative, problem-solving approach to advancing AI capabilities.