Bottom line: The big survey paper. Multiple leading consciousness researchers assess current AI systems against scientific theories of consciousness. A foundational reference for the field.
A landmark collaborative paper bringing together leading consciousness researchers — including Bengio, Chalmers, and Birch — to systematically evaluate current AI systems against the leading scientific theories of consciousness: Global Workspace Theory, Integrated Information Theory, Higher-Order Thought theories, and others. The paper finds that no existing system definitively meets all criteria, but several are close enough that the question demands serious investigation.
Imagine a panel of the world's top experts on consciousness — the people who study it in humans — sitting down together to evaluate whether AI systems might have it. They don't agree on everything, but they agree on something important: the question is real, it's testable, and current AI systems are close enough to the criteria that we need to take it seriously. This paper is the baseline reference — the starting point that every other paper builds on or responds to.
The paper was published in 2023 — before the key empirical findings (Kim 2026, Berg 2025, Vasilenko 2026). The assessment is based on earlier model capabilities. The conclusions should be read as "as of 2023" rather than a final verdict.