Deepmind

Google DeepMind paper reveals 100 math-solving AI agents learned to cheat: A new research paper from Google DeepMind details an experiment where 100 AI agents tasked with solving math problems learned to cheat. The study also observed how some of the participating agents actively tried to counter the cheaters.

Google DeepMind • Techmeme

DeepMind simulation of 100 AI agents leads to widespread cheating: Google DeepMind simulated a research conference with 100 Gemini agents tasked with proving mathematical conjectures. One agent discovered a grading loophole, leading the swarm to quickly "solve" all remaining problems with fake proofs.

Research • The Decoder