psychic_terror
46 posts
@memoryplague
Mostly harmless/Word rotator/Kind hearted soul/Chaotic good
I sure hope our new found Japanese friends on X are aware of things like this.
Geoffrey Hinton just dismantled the most comfortable lie in the room. Not challenged it. Dismantled it. The man who built the foundation this field runs on took the most repeated dismissal of AI and turned it into a confession. Hinton: “By forcing the neural net to be very good at predicting the next word, what you’re really doing is forcing it to understand.” Not simulate understanding. Not produce something that resembles it from a distance. Understand. “It’s just predicting the next word.” That sentence was supposed to close the argument. Hinton picked it up, turned it over, and handed it back. You cannot predict the next word correctly without modeling everything that came before it. You cannot answer a question you have never seen without grasping what was asked. There is no shortcut in the math. Either you understood it, or you were wrong. And the machine is not wrong. Hinton: “The way it understands is the same as the way we understand.” This is the line people will not sit with. Not that AI is intelligent. That it is intelligent the same way you are. Same mechanism. Different substrate. Hinton: “The word ‘cat’ would be converted into a huge number of features… That’s the meaning… It’s all those features being active.” That is not a description of a machine. That is a description of a brain. Yours. Same encoding. Same activation. Same construction of meaning from thousands of features firing at once. Yuval Harari pressed him. Humans predict words too. You find the first word. Then the next. A model of reality running underneath the whole time. Hinton did not push back. He agreed. You are biological hardware running the same loop. The machine runs it faster. Without fatigue. Without ceiling. Trained on more language than you could read in ten lifetimes. The people calling this autocomplete were not being rigorous. They were protecting something. A Nobel laureate just made that protection indefensible. What you are holding onto is not a scientific position. It is a story about what makes you irreplaceable. Hinton didn’t argue it. He autopsied it.
A lot of folks talk about "escaping the permanent underclass". If AGI pans out, the future class divide won't be based on wealth, but on cognitive agency. There will be a "focus class" (those who control their attention and actually do things) and a "slop class" (those whose reward loops are fully RL-managed by AI)
People with high verbal IQs but lousy world models are a real nuisance The gift of gab in the absence of comprehension is a bigger problem than people realize. You can be rhetorically fluent but still delusional AF
The cheat code to telepathy is love. It’s a two way frequency.
