Geoffrey Hinton is Wrong About AI Understanding

Do AI models actually "get it," or are they just guessing the next word?

In this video, we dive into the philosophical and technical debate over whether Large Language Models (LLMs) truly possess understanding. While many see AI as a sophisticated statistical tool, "Godfather of AI" Geoffrey Hinton argues that the internal structures these models build constitute genuine understanding.

In this video, we explore:

• Defining Understanding: What does it mean for a human to understand a task vs. a machine?
• The Reality Gap: Why being trained on text alone might mean LLMs lack a "grasp" of the physical world.
• The Hinton Perspective: An analysis of Hinton's claim that internal data structures are the definition of understanding.
• The Rebuttal: Why Hinton’s definition might be missing the vital connection between mind and reality.

Is understanding a structural phenomenon, or does it require a pulse and a place in the real world? Let’s get into it.