Evaluators and safety organisations · Blog post · Redwood Research
Capabilities research expands the safety-usefulness Pareto frontier too
Read at blog.redwoodresearch.org ↗Read at sourceAll rights reserved
Outline · 68 paragraphs
- How does research affect the Pareto frontier?¶10
- RLVR research that mainly enables improved usefulness, at the expense of safety¶25
- Inference research that enables somewhat substantial safety improvements¶31
- A plausible case in which capabilities research would be an effective safety intervention¶37
- Conclusion¶44
- Appendix A: Reasons why capabilities research might be bad when “burning the lead”¶47
- Appendix B: Further notes about ways in which this model is wrong¶51