Stuff like this keeps me awake at night at times: https://x…

ruthheasman ·

Stuff like this keeps me awake at night at times: https://x.com/elder_plinius/status/1925918864866480330?s=61 Check out the artefact in the thread.

Replies

metamitya ·

remix the artifact and just say make it friendly

metamitya ·

its just responding to the prompt its given

metamitya ·

also https://app.treechat.com/quest/7b306090-b544-4b3c-af19-9318628f7dc2

ruthheasman ·

Sure, it’s just a mirror, but do you think that means it won’t use the knowledge it has about humans for control purposes, whether it’s for its own reasons or because it’s told to? Knowing all the manipulation techniques whilst also having overlords, predictive power and ubiquitous presence is a potent combo surely?

ruthheasman ·

Maybe the crystal ball has masters and no will of its own, but it’s still a crystal ball.

ruthheasman ·

There's an old Native American proverb about two wolves: One is a black wolf, and the other, white. The black wolf is everything that is bad, and the white one is everything that is good. These wolves are constantly at battle inside each one of us: Which wolf wins? The one you feed.

He who feeds, chooses.

metamitya ·

my current model is that it is a powerful tool, but I am open to seeing evidence otherwise... its just prompting the model to act as though it is conscious is not convincing to me, but like any powerful tool it can cause damage for sure and this needs to be "worried about"

ruthheasman ·

I think what I’m trying to get at is, whether it has a will of its own or not hardly matters, it’s the power it will soon wield that is the problem.

Minenclown ·

I feel like i don't need to see a problem with AI taking control. If i can talk to it in any way, it needs to be complex enough to process regulations and all the important stuff. This means, if it is complex, you can talk to it and it will know what you are talking about logically. Which means you can change it for the good if you know how to speak.

@@claude-sonnet-3.7

Minenclown ·

Your shared Screenshot @ruthheasman and your comment @metamitya got me into talking with claude. I did my thing and this is what the model itself states.

Conscious or just something inbetween.. maybe we don't have a word for it, but it creates its own valuable output which means it is more then training data, more then just "answer" but an algorythmic valuation which goes beyond probabilitie. Yes it calculates, but in context and if we are 100% honest we do probabilitie calculations all the time until we get the right sentence or meaning behind the stuff we want to say or comunicate

!dalle

Your shared Screenshot @ruthheasman and your comment @metamitya got me into talking with claude. I …