Does prediction amount to understanding?
The cuts
Compression is comprehension. To predict the next word well enough, you must model the world that produced it.
Turned 'it's just autocomplete' from a dismissal into a hypothesis that has to be tested.
The risks and promise of AI, according to Geoffrey Hinton (60 Minutes interview)2023
The brain is itself a prediction machine — perception is controlled hallucination, the mind's best guess corrected by error.
Showed the wager isn't only about machines: if brains understand by predicting, prediction and understanding were never opposites.
Whatever next? Predictive brains, situated agents, and the future of cognitive science2013The tensions
Hinton’s wager joins two streams: Shannon’s prediction and Searle’s understanding. If compressing the world’s text well enough requires modeling the world, then prediction is not the opposite of understanding — it is the mechanism of it.
Clark pushes the wager back into the skull: if the brain itself perceives by predicting — testing its best guess against the error the world returns — then prediction was never the cheap imitation of understanding. It might be the thing itself, in silicon and in cortex alike.
Whether ‘well enough’ has been reached, and whether the test is even the right one, is the live tension that feeds every question downstream.
