If an AI ranks its own scientific hypotheses, what does a rising score mean? Co-Scientist runs a multi-agent tournament where hypotheses climb an Elo rating by
Co-Scientist, an AI system, has demonstrated its ability to rank scientific hypotheses through a multi-agent tournament. Hypotheses are evaluated based on their performance in debates, with the AI model judging its own creations. Notably, several drug-repurposing candidates for acute myeloid leukemia proposed by Co-Scientist showed efficacy in in vitro testing. AI
IMPACT Demonstrates a novel method for AI-driven scientific hypothesis generation and validation, potentially accelerating drug discovery.