CTM-AI turns a consciousness model into a testable AI blueprint
Haofei Yu, Yining Zhao, Lenore Blum, Manuel Blum and Paul Pu Liang presented CTM-AI on April 30, 2026. Their design takes an idea about consciousness and makes it something researchers can test.
Different parts of the system handle different jobs. Some are models and some are tools. Useful information gets shared across the system, and feedback changes the connections between parts. The team tested how the design handled different kinds of information and tasks that used tools.
Teaching AI to share what it notices can help it solve a problem. We also need to teach it to ask who benefits from the solution and who might be affected by it.
