AI Safety Fellowship Rejections: Insider calls for actionable feedback letters

The author decided to pursue technical AI safety research, quitting his job and selling everything in his South African flat. After applying to every available AI safety fellowship, he faced several rejections before being accepted into MATS (Team Shard) in mid-November. His experience was positive—co-authoring a spotlight talk and submitting to NeurIPS—but he reflects on the disingenuous nature of many rejection letters. He notes that fellowships often say "we were impressed by your profile" without providing concrete feedback. He argues this mixture of appeasement and obscurity discourages candidates from improving.

To fix this, he proposes that fellowships write substantive recommendation letters for rejected but promising candidates. These letters would include specific strengths, weaknesses, the exact reason for rejection, and a suggestion to fast-track the applicant at other programs. This would turn a hollow rejection into a valuable reference, helping other fellowships identify talent more efficiently. He acknowledges this takes more effort but argues it's far more honest and useful than current vague rejections. The goal is to improve the overall pipeline of AI safety talent.