Science of machine consciousness
A validated theory and test battery that can tell whether an artificial system has subjective experience, and what architecture would produce it deliberately.
State of the art
Butlin, Long et al. indicator framework (2023; Trends in Cognitive Sciences 2025) concludes no current AI is clearly conscious and no obvious barrier exists; no validated test.
- Competing theories (global workspace, IIT, higher-order) make conflicting predictions
- Behavioral reports are trivially imitated by language models
- No ground truth for consciousness outside humans to calibrate tests
Who’s working on it
- Patrick ButlinResearcherUniversity of Oxford
Lead author of the 2023 indicator-properties report deriving consciousness markers from competing neuroscience theories, and of Conscium's 2025 research principles
- Robert LongResearcherEleos AI Research
Directs Eleos and co-authored Taking AI Welfare Seriously (2024) with Chalmers, Birch and Sebo, which set the current research agenda
- Eleos AI Research ↗Program
Nonprofit founded 2024-2025 focused on AI moral patienthood; published the 2025 priorities for welfare-relevant evaluations and introspection tests
- COGITATE ↗Program
Templeton-funded adversarial collaboration led by Lucia Melloni testing Integrated Information Theory against Global Neuronal Workspace head-to-head with fMRI, MEG and iEEG
- Lucia MelloniResearcherMax Planck Institute for Empirical Aesthetics
Leads COGITATE, the first well-powered preregistered attempt to decide between the two leading theories rather than confirm either
- Kyle FishResearcherAnthropic
Anthropic's first model-welfare researcher (hired 2025); ran pre-deployment welfare assessments on Claude 4 and has published explicit credence ranges on model consciousness
- Conscium ↗Startup
UK company that published five principles for machine-consciousness research with Butlin in 2025 and gathered 169 signatories, the only commercial actor working on this directly
What to invest in
| Kind | Investment | Scale |
|---|---|---|
| research | Adversarial collaborations testing rival theories in humans and animals, then transferring tests to AI | $$ $10M – $100M |
| research | Interpretability-based audits for theory-derived indicator properties in frontier models | $ < $10M |
Depends on
Required by
No other capability builds on this directly.