Comment by int_19h
4 years ago
> it basically just regurgitates the majority opinions that can be found on the internet on those topics (sometimes combining several of those opinions in an incoherent way)
And this is different from your average Facebook comment how exactly?
I wonder just how much our refusal to acknowledge these things is because it would be a tacit admission of how much we've overestimated our own intellect as a species. What if we ourselves are the "not really intelligent" model with delusions of grandeur, and that is why the AI can already achieve so much? What if most of our civilization does not, in fact, require "real intelligence" to function?
It is very different from having a discussion with another human, following their reasoning, and them following your reasoning, and in that way exploring, developing, validating and invalidating ideas in a dialog. This is one application I would find an AI useful for, as a “second brain” to bounce ideas off of, or to get it to explain things to me that I need clarification on, and it understanding the specific objections or gaps I try to explain back to it.
I think the main differences are internal state and feedback loops. The language model is stateless and immutable, whereas human brains are stateful and plastic.
To mimic state within ChatGPT all the previous prompts and responses are fed together with the last one. But of course, that's not very efficient and quite limited.
While it's certainly impressive right now, it's nowhere close to general intelligence. Once AI will be able to keep track of stage and update itself through feedback loops, then I'd say we'll be pretty close. Right now it's a language model that is extremely good at detecting patterns and applying them.