Lilian Weng on harness engineering for AI self-improvement

Explores harness engineering as a method to control recursive self-improvement in AI systems. References I.J. Good's 1965 concept of an ultraintelligent machine and Yudkowsky's 2008 work on recursive improvement.