OpenAI CFO cites recursive self-improvement as cost-saving; AI researchers continue despite existential risks

OpenAI CFO Sarah Friar stated that the company's largest models can now train smaller models, reducing training costs—an example of recursive self-improvement (RSI). This development highlights the ongoing pursuit of AI capabilities despite researchers' stated beliefs that such systems could pose existential risks. The claim is based on a Business Insider report, and the broader context of AI safety concerns remains unresolved.