Today while running, I was pondering how human intelligence improves and is passed down through generations. After much thought, I realized that the core aspects are exploration and distillation.

Exploration helps humans find new hypotheses, which combined with feedback generate new knowledge. For example, when Yang Zhenning and Li Zhengdao proposed that weak interactions violate parity conservation, and Wu Jianxuan’s experiments provided feedback, this process created new physical knowledge.

Distillation is how existing knowledge is transmitted to the next generation, such as teachers passing knowledge to students. Moreover, students can distill knowledge from multiple teachers, enabling them to surpass each individual teacher.

To create AI that exceeds human capabilities, we could employ similar methods.

(This post is translated from my original post in Chinese)