“Consciousness in Artificial Intelligence: Insights from the Science of Consciousness” is a 2023 report by a large interdisciplinary group, including Patrick Butlin, Robert Long, Yoshua Bengio, and philosophers of mind such as Jonathan Birch. It tackles a question usually left to speculation - could an AI system be conscious - and tries to make it empirical rather than rhetorical. The authors deliberately set aside the deeper philosophical puzzle of why experience exists and instead ask a more tractable question: do AI systems have the functional features that leading scientific theories associate with consciousness in brains.
The method is to survey prominent neuroscientific theories - recurrent processing theory, global workspace theory, higher-order theories, predictive processing, and attention schema theory - and translate each into “indicator properties,” concrete computational characteristics that can in principle be checked in an AI system. The approach assumes computational functionalism: that consciousness depends on the right kind of information processing rather than on biological material specifically. The report then asks whether existing systems exhibit those indicators.
Its headline conclusion is that no current AI systems are conscious. But it adds a pointed caveat: there appear to be no obvious technical barriers to building systems that would satisfy many of the indicators. In other words, the authors find that nothing they identified is uniquely tied to biology, so the question is likely to recur as architectures grow more brain-like. The report is widely cited as a careful, theory-grounded alternative to both casual claims that chatbots are sentient and flat dismissals that the question is meaningless.