A state that tracks entities and their relations over time could represent a changing situation. Neural network representations, by contrast, are recomputed from current inputs and do not persist as an independent world model. section 5.4.5
Figure 5.5 Human language refers to events and objects in the world through concepts grounded in perception, action, and consequence. Artificial neural networks instead learn patterns from indirect records of the world, such as text, images, or other data. As a result, models can generate coherent outputs by connecting representations within those records, without attaching them to a world they directly experience.

The figure contrasts human concepts grounded in perception and action with patterns learned from indirect records such as text and images. section 5.4.5



















































