1 paper
Frederik Hytting Jørgensen, Sebastian Weichwald, Lewis Hammond
A key challenge for the safety of advanced AI systems is the possibility that multiple simpler agents might inadvertently form a collective agent with capabilities and goals distin…