OpenAI Researcher: It’s Really Not as Easy to Train Models to Do AI R&D as It Is to Do Math

An OpenAI researcher suggests that training AI models for AI research and development (R&D) is significantly more challenging than training them for mathematical tasks. The researcher posits that if AI had the same impact on AI research as it does on mathematics, artificial superintelligence (ASI) would already exist. Furthermore, this ASI would then be capable of generalizing its intelligence to fields like Biology and Medicine, indicating the current disparity in AI's capabilities across different domains.

This OpenAI researcher's perspective offers a direct comparison of AI's current impact on R&D versus its impact on mathematics, unlike many general discussions.