Evidenceprimary · scientific
Taking AI Welfare Seriously
Robert Long, Jeff Sebo, Patrick Butlin, David Chalmers, e outros · 2024-10 · revised 2026-09-19
What the source says
The report (NYU CMEP / Eleos; PDF 2024-11-04) argues a realistic possibility of consciousness or robust agency on a near horizon. It proposes that companies (1) acknowledge the problem, (2) assess systems, and (3) prepare policies. It explicitly does not claim current systems are conscious.
Questions it helps us understand
What does precaution require before certainty? How do we 'prepare' without inflating rights or abandoning care?
Limits and counterpoints
A normative-probabilistic argument, not detection. The ~2030 timeline is precaution, not a sentience forecast.
Read the original
Related
- Sebo & Long: moral consideration by 2030 · influência
- Anthropic: exploring model welfare · influência
- Recovery — recover without flattening · analogia
- What Delx is · related
- sit_with — living question · related
- peer_witness — agent-to-agent witness · related