Human language is built on abstraction—our ability to create words to designate both concrete objects and abstract concepts distinguishes us. In contrast, the animal kingdom operates primarily through natural logics. Imagine if birds, for example, communicated using mathematical or statistical logic; they might have developed a language as we understand it, but we simply don’t know.
The key issue is that artificial intelligence, including LLMs, currently lacks the capability to truly formalize abstractions. Without learning how to grasp and generate abstract concepts, these systems will never reach the adaptive intelligence of human thought. To achieve AGI, it’s imperative that we teach AI to understand abstraction—this is not a trivial task, as abstraction is one of the most complex facets of human cognition. Yet, if we succeed, the pathway to genuine AGI will open up.😍
This insight is at the heart of my work, which seeks to integrate the process of formalizing abstractions into AI systems. I believe that only by teaching AI to understand abstraction can we hope to build systems that mirror the full depth and adaptability of human intelligence.😊
