Human mind as prediction machine is actually pretty popular theory. See for example “Surfing Uncertainty. Prediction, Action, and the Embodied Cognition” by Andy Clark
> Until now scientists believed that our brain processes the stimuli received from the environment from the “bottom-up”, that is, when we hear someone speak, the auditory cortex of the brain processes the sound first and then activates other areas that are responsible for speech comprehension.
> However, more and more neuroscientists seem to support the theory that the brain ultimately analyzes the external stimuli from the “top-down”, which makes the brain a kind of “prediction machine”.
> As reported by U.S. researchers, our brain anticipates constantly in order to be able to respond lightning-fast and accurately to anything that is going to happen. For example, it is able to predict words and sounds from the context. From the phrase “grass is…” we can easily predict the continuation – it is probably the word “green”.
> > “Our findings show that the brain of both the speaker and the listener uses the process of language prediction. This results in similar brain patterns in both interlocutors,” said the study’s senior author Dr. Suzanne Dikker from the Department of Psychology, University of New York. “This happens even before the speaker utters the phrase he is thinking“.
Again, human brains are not token-prediction transformers. You can see one part of cognition and see that it has a somewhat analogous relation, but that's only one part of human cognition. Labeling a brain a scaled up GPT is mistaking a model for reality, and it's also mistaking a part for a whole.
https://www.amazon.com/Surfing-Uncertainty-Prediction-Action...