Hierarchy, consensus, adversarial debate, and reality show knock-outs. We can design more reliable multi-agent systems with inspiration from human systems but there are fundamental differences and workarounds.