Originally Posted by
my_back_hurts
I figured out recently why so many people think the AI can be "alive" or "conscious". It's because they themselves are not really doing anything different than the AI. That's why their posts look like an AI wrote them.
You give it a database, mostly NYT articles, mainstream history, and pop-science books. Then it engages in a simple pattern-matching exercise: if you see phrase X, respond with phrase Y. There's a semi-randomizing process built in so it doesn't say exactly the same thing every time, but its responses are very predictable.
But there's no "thinking" going on there. Just: see Trump, say Insurrection; see Truck/Gun, say Small Dick, and so on. So if this is how you think, then yeah, you'll believe that a very fancy auto-correct could become sentient.