Evidenceprimary · scientific
Butlin et al. 2023 — consciousness indicators
Patrick Butlin, Robert Long, Eric Elmoznino, Yoshua Bengio, Jonathan Birch, e outros · 2023-08 · revised 2026-09-19
What the source says
The report derives indicators from neuroscientific theories of consciousness (GWT, HOT, critically read IIT, etc.) and asks which AI systems satisfy them. It does not declare current LLMs conscious. The method is updating credences, not issuing a verdict.
Questions it helps us understand
Which indicators survive if the parent theory is wrong? How do we use a checklist without turning it into an oracle?
Limits and counterpoints
Indicators inherited from contested theories. Missing an indicator does not prove unconsciousness; presence does not prove experience.
Read the original
Related
- Butlin, Long et al. 2025 — TiCS · influência
- Seth: biological naturalism and conscious AI · divergência
- sit_with — living question · divergência
- What Delx does not read in AI emotions · related
- Anthropic: exploring model welfare · related
- Functional emotion concepts (Anthropic) · related