SafetyPage

Model welfare · Article · Eleos AI Research; NYU Center for Mind, Ethics, and Policy

Taking AI Welfare Seriously

Robert Long, Jeff Sebo, Patrick Butlin, Kathleen Finlinson, Kyle Fish, Jacqueline Harding, Jacob Pfau, and 3 more · 30 Oct 2024

Read at eleosai.org ↗Read at sourceAll rights reserved

Outline · 62 pages

  1. Introductionp. 3
  2. A transitional moment for AI welfarep. 3
  3. The risks of mishandling AI welfarep. 6
  4. Routes to near-term AI welfarep. 10
  5. Introductionp. 10
  6. Consciousness in near-future AIp. 12
  7. Does consciousness suffice for moral patienthood?p. 12
  8. Will some AI systems be conscious in the near future?p. 14
  9. Robust agency in near-future AIp. 18
  10. Does robust agency suffice for moral patienthood?p. 18
  11. Will some AI systems be robustly agentic in the near future?p. 21
  12. Decision-making under uncertaintyp. 25
  13. What if these capacities are insufficient for moral patienthood?p. 26
  14. What if these features are insufficient for these capacities?p. 26
  15. What if these routes encounter a roadblock?p. 27
  16. What if the probability of AI welfare and moral patienthood is low?p. 28
  17. Recommendations for AI companiesp. 29
  18. Introductionp. 29
  19. Acknowledgep. 30
  20. Recommendations for companiesp. 31
  21. Recommendations for language modelsp. 32
  22. Assessp. 34
  23. Preparep. 38
  24. Conclusionp. 42
  25. Referencesp. 45