> Put the LLM/agent/AGI/SI in a box with full access to the internet. Don't give it any instructions. Nothing. Not even a system prompt. What happens?

This doesn't even make sense. Without a system prompt, how will it know that it has tool access?

Put man in pitch-black room, why doesn't he read book or talk to me?

▲

pj_mukh 13 hours ago | parent | next [-]

Actually man freaks out and starts hitting the walls.

But either way, setting that asides by “give it full access to the internet” I meant giving it the tools to access the internet.

It just won’t though. It has no need to.

▲

int_19h 12 hours ago | parent [-]

The weights themselves are inert. It would be like complaining that a brain preserved in formaldehyde isn't thinking.

What you need is a loop. And if you try running an LLM in a loop with internet access, it will eventually use it.

	▲	pj_mukh 6 hours ago \| parent [-]
		What does an LLM with zero instructions do nowadays with no loop? It's basically just noise spilling from the weights and "snapping" to the closest actions/tools , with no real signal going through?

▲

Applejinx 12 hours ago | parent | prev [-]

But that's exactly it. It doesn't 'know'. It isn't. It's a possible story nobody told. There is nothing there.

▲

peddling-brink 8 hours ago | parent [-]

Your brain cut off from any stimulus doesn't "know" either. There are a lot of reasonable arguments to be made against AI, but this isn't one of them.

Tool bad because I used it wrong.

	▲	willy_k 7 hours ago \| parent [-]
		“Know” isn’t a perfectly useful word here. The mind still experiences sans material stimulus, for example dreaming. And it experiences in a progressively less goal/immediate-task-relevant-stimulus manner as stimuli go down (default mode network activity, seen in dreaming, sensory deprivation etc) - i.e. the mind experiences, and processes (dreams about taxes, and so on) independent of the world.