They pitted humans working alone against humans collaborating with large language models (LLMs), specifically OpenAI's GPT-4o, and memes generated entirely by GPT-4o without human input.