The Prediction Engine in Your Skull

Watch yourself finish someone else's sentence and you catch your own machinery working. The next word arrives before you decide on it. Prediction is not a feature bolted onto thinking, it is most of what thinking is doing.

Ore's point is not that a language model is a brain. The architectures are not the same and he does not claim they are. The point is narrower and harder to dismiss: the thing people find alien about next-token prediction is a thing they do continuously and have never watched themselves do.

That reframing changes the emotional register of the whole debate. An alien intelligence is something to be defended against. A mirror is something to be used. He argues the second framing is both more accurate and more useful, and that the fear in the first one is mostly unexamined.

The behavioral read matters here. People assess novel technology through availability and threat, not through mechanism, which is why the conversation reliably lands on replacement rather than on capability. Naming the mechanism is what moves it.

The essay closes where his work does. If the machine is a mirror and possibly a partner, the only question worth spending time on is what you point it at.