WHY WOULD AI COMPANIES USE HUMAN-LEVEL AI TO DO ALIGNMENT RESEARCH?

Summary from the TLDR AI newsletter

AI companies might not prioritize alignment bootstrapping when they have human-level AI, mirroring their current focus on advancing capabilities with human researchers over safety efforts. This trend could lead to a dangerous disparity where AI capabilities significantly outpace alignment, increasing existential risks. To avert this, AI companies should already be emphasizing safety over capabilities to prove their commitment to responsible development.