Figure 1.

Comparison of conventional AI and hypothesis-driven AI. (A) The learning pipeline of conventional AI. The learning algorithms often include weighted connections and distance metrics without the need to include existing domain knowledge or a hypothesis into the design of learning algorithms. (B) The design and learning pipeline of hypothesis-driven AI. Knowledge or hypothesis are the built-in components in the design of learning algorithms; these facilitate the discovery of novel associations between attributes that can explain data properties.