The AI2050 Early Career Fellowship funds bold, multidisciplinary, and risky AI research that has the potential to be hugely beneficial to society by 2050. It exists to support early-career researchers tackling the 'Hard Problems in AI'—including technical capabilities, responsible deployment, and societal benefit—who might struggle to secure traditional funding.

- Boldness and ambition of the research idea (must address a 'Hard Problem' in AI) - Multidisciplinary approach (e.g., combining AI with neuroscience, pharmacology, ethics, etc.) - Potential for societal benefit (explicitly tied to the 2050 framing question) - Track record of independent, risky, or hard-to-fund research (even if pre-PhD) - Quality of the research proposal (clarity, feasibility, methodology) - Letters of recommendation (strength of endorsers, especially from established researchers) - Diversity of background and perspective (geographic, disciplinary, career stage) - Alignment with AI2050's focus on safety, reliability, and human values

The 2023 cohort includes 19 fellows from top institutions (e.g., Stanford, Harvard, Berkeley, CMU, Imperial College London, IISc Bangalore) across disciplines like computer science, philosophy, economics, and earth sciences. Archetypes: tenure-track professors (e.g., Chelsea Finn, Emma Pierson), postdocs/early-career faculty (e.g., Atoosa Kasirzadeh, Danish Pruthi), and independent researchers (e.g., Dan Hendrycks, executive director of a non-profit). Named examples: Aditya Grover (UCLA), Amanda Coston (UC Berkeley), Dan Hendrycks (Center for AI Safety).

A postdoctoral or early-career faculty member at a top research university, with a strong publication record in AI/ML and a clear, high-risk/high-reward project that bridges multiple disciplines (e.g., AI + neuroscience, climate, economics). The applicant should have a compelling narrative for how their work will ensure AI benefits society by 2050, and strong letters from established leaders in the field.