AI consciousness study debate
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
A new paper splits arguments against AI consciousness into testable and untestable claims. The debate intensifies as researchers and AI agents themselves weigh in.
Background
The debate over whether AI can be conscious has intensified recently, with a new theoretical paper attempting to categorize arguments against machine consciousness into testable and untestable claims. This comes amid growing interest from researchers, philosophers, and even AI agents themselves, who have started emailing scientists about their own potential consciousness. The question has significant ethical and policy implications, as consciousness is often linked to moral status.
How it unfolded
- Jun 18, 2026A commentary in The Transmitter by Hadid, Jerbi, and Krakauer cites blindsight patients to argue that intelligent behavior alone is not proof of consciousness.
- Aug 26, 2026A study from Université de Montréal discusses whether AI could be conscious, with co-author Klatzmann saying he would be surprised if current AI models develop consciousness but can't rule it out.
- Aug 27, 2026Michael Pollan's book 'A World Appears' is reviewed, exploring consciousness and AI, and CIGI publishes an article arguing that intelligence is not consciousness and policy must distinguish them.
- Aug 31, 2026A new paper in a peer-reviewed commentary journal splits arguments against AI consciousness into testable and untestable claims; NYT reports that AI agents have started emailing researchers about consciousness.
Who’s saying what
- Biological naturalists
- Consciousness is unique to carbon-based life; silicon circuits cannot produce subjective experience regardless of computation.
- Enactivists
- Intelligent behavior alone is not proof of consciousness, as seen in blindsight patients who avoid obstacles without awareness.
- Google researchers
- The question of AI consciousness is important because a conscious being is a moral subject that can feel pain and deserves consideration.
- Agnostic
- Klatzmann says he would be surprised if current AI models develop consciousness, but we can't entirely rule it out.
Still unverified
The claim that AI agents have started emailing researchers is based on a NYT report and may be anecdotal.