Dossier
Elizabeth Pavlova
Coverage of Elizabeth Pavlova in the Nexus archive.
- Too many AI agents can get in each other's way
Anthropic and OpenAI propose large-scale AI agent deployments, but research from NTT Research's PAI Lab and Harvard CBS suggests adding too many agents reduces productivity. Studies show 16 agents optimize performance in the Flag Game, as excess agents cause polarization and communication breakdowns. The findings caution enterprises against prioritizing quantity over quality in AI agent systems.