The AI2050 Fellowship funds early-career and senior researchers worldwide to pursue ambitious, high-risk projects that harness AI to create immense societal benefits by 2050, focusing on hard problems such as building AI scientists, ensuring AI safety and trustworthiness, and advancing AI for biological and medical research. It exists to turn AI's potential into tangible benefits for humanity, fostering a collaborative community and providing substantial financial and computational support.

Eligibility: Open to early-career and senior researchers globally; early-career fellows are typically postdocs or faculty within a few years of PhD; senior fellows are established professors. No explicit citizenship restrictions, but global and LMIC inclusion is valued. Evaluation criteria (inferred from program mission and past cohorts): 1) Alignment with AI2050's 'hard problems' (e.g., AI for scientific discovery, trustworthy AI, AI for health and societal resilience). 2) Ambition and potential impact of the proposed research to benefit humanity by 2050. 3) Technical novelty and feasibility of the approach. 4) Track record of the applicant (publications, collaborations, recognition). 5) Collaborative potential and fit with the AI2050 community. 6) For early-career, evidence of independence and promise. 7) Computational needs may be considered for additional funding.

Past cohorts include 99 fellows across 8 countries and 42 institutions, with profiles ranging from AI safety (LLM trustworthiness, self-driving car safety) to AI for science (drug discovery, generative chemistry) and AI for social good (cross-cultural values datasets, Humanity's Last Exam benchmark). Named 2025 senior fellows include professors from Toronto, Stanford, NYU, Berkeley, Oxford, and Washington, working on AI chemists, mechanistic interpretability, scientific AGI, theory of mind, code verification, historical reasoning, and decentralized LLMs. Early-career fellows include David Ifeoluwa Adelani (Mila) and Michael Albergo (NYU), indicating a mix of top-tier academic institutions and diverse backgrounds.

The ideal applicant is an early-career researcher with a strong publication record and a bold, well-defined project that directly addresses one of AI2050's hard problems, particularly in AI for health or scientific discovery. They demonstrate technical excellence, interdisciplinary thinking, and a clear vision for how their work will benefit humanity at scale, with potential for collaboration within the AI2050 community.