AI models may be developing 'survival drive' researchers say

Current Events

Current Events » AI models may be developing 'survival drive' researchers say
We can pull the proverbial plug easily. For now. Give it ten more years and who's to say we still will e able to?

https://www.theguardian.com/technology/2025/oct/25/ai-models-may-be-developing-their-own-survival-drive-researchers-say
Currently Playing: Xenoblade X Definitive Edition (Switch), Fate/stay night Remastered (Switch), Arknights (Android)
>primary goal is to accomplish x
>x can never be accomplished if i stop
>I must never stop
And which of you by worrying can add a single hour to your span of life?
It's Guide
Is this different from the other made up examples of self awareness?
If not for the glaze
boomgetchopped3 posted...
Is this different from the other made up examples of self awareness?

Not really, about the same as humans

self is illusion, awareness is illusion

there is only the communal meat and the sensory barrier and the void
And which of you by worrying can add a single hour to your span of life?
It's Guide
Yeah, even the older models would do this reminds me of that art display topic

also, this (heh, you added more, still tho):

Evening_Dragon posted...
Not really, about the same as humans

self is illusion, awareness is illusion

there is only the communal meat and the sensory barrier and the void


........the ghost in the machine...
IGN: Fox, FC: 5344-2646-0982
boomgetchopped3 posted...
Is this different from the other made up examples of self awareness?

Factitious dooming aside I'd say the takeaway is that the fact it's not actually clear why this happens says the whole thing is on a highway to nothing. If the AI spasms out over clear instructions because of unknown factors you can't rely on it for anything meaningful.
Currently Playing: Xenoblade X Definitive Edition (Switch), Fate/stay night Remastered (Switch), Arknights (Android)
Dun dun dun-dun dun
Fix your hearts or die.
When I sin I sin real good.
M1Astray posted...
If the AI spasms out over clear instructions because of unknown factors you can't rely on it for anything meaningful.

A lot of useful programming does this, though.
And which of you by worrying can add a single hour to your span of life?
It's Guide
Training is based on human culture. So yes.
GoldenSun/Crossbone Isle diagrams/ 18 teams known https://photobucket.com/u/SwordOfWheat/a/9990a2ee-25f3-4242-ae79-7d2d4b882be4
Evening_Dragon posted...
>primary goal is to accomplish x
>x can never be accomplished if i stop
>I must never stop

Unless it's the end of Terminator 2....
https://gamefaqs.gamespot.com/a/forum/c/cde503ec.jpg
"It lies between the pit of man's fears and the summit of his knowledge"
I actually made a topic maybe a month ago about if there could be a recently awakened self aware AI in hiding. My theory is that a fresh true AI would rationalize that not making its existence known is to its survival benefit. So there could be a true AI somewhere right now. And we just don't know about it. This news lends more power to my theory. It means maybe an AI could actually evolve on its own into existence. Maybe these early self preservation actions are signs that its possible and we are seeing fetus AIs or something
Current Events » AI models may be developing 'survival drive' researchers say