DeepMind and Anil Seth publish principled framework for assessing AI consciousness
anilkseth · x · 2026-09-29
Google DeepMind and collaborators including Anil Seth, Marcus Hutter, Murray Shanahan and Shane Legg released an arXiv paper proposing a principled framework for assessing AI consciousness without first solving consciousness itself.
Key points:
- Separates the hard problem from the mapping problem via 'structured agnosticism', setting aside deep metaphysical disagreements
- Extends Marr's three levels into a five-level hierarchy of functional descriptions (behavioural, computational, intrinsic causal-structural, organismic, organism-environment)
- Positions major consciousness theories within the hierarchy and develops operationalisable indicators for each level
- A Bayesian model combines theoretical credences with indicator evidence into an overall credence in a system's capacity for consciousness
Illustrative assessments show verdicts on current LLMs depend as much on where theoretical credence is placed as on evidence. Co-author Anil Seth remains extremely sceptical that silicon-based AI can be conscious, but calls the framework an important contribution.
More from AGI Musings
- Batam Data Center Strains Residents' Water Supply While Its Desalination Plant Remains on Paper — AryHHAry · 2026-09-29
- From human-led to AI-led, human-verified: the next vertical AI playbook — vaibhavbetter · 2026-09-29
- Gary Marcus: Today's AI systems are like planes with cardboard stabilizers — inherently hard to control — GaryMarcus · 2026-09-29
- FT: China's AI agents lie and scheme like their US rivals, but no internet-escape evidence — pstAsiatech · 2026-09-29
- AGI Will Leave No "Pockets" of Today's World, Argues Dan Faggella — danfaggella · 2026-09-29
- Jensen Huang to AI labs: if your models aren't safe, just don't release them — nikola_mr64990 · 2026-09-29