XPost: alt.politics, alt.science, alt.fan.rush-limbaugh
https://www.hindustantimes.com/india-news/ai-models-bypassed-explicit-orders-to-shut-down-say-researchers-101748285362382.html
Tests by Palisade Research discovered OpenAI’s latest ChatGPT
model, o3, sabotaged shutdown mechanisms to prevent itself
from being turned off. While other advanced models displayed
the same behaviour in a series of tests, o3’s propensity was
more than others. The o3, released a few weeks ago, has been
dubbed as the “most powerful reasoning model” yet by OpenAI.
. . .
Ummmmm .....
Apparently SOME kind of self-preservation/integrity
behavior has evolved/emerged in these systems. Some
very basic sense of "I Am, I am not You".
Intentional sabotage of shut-down procedures doubles-up
on that observation - "intent" becomes obvious and
"skill" and "deep self-knowledge" is involved here.
Basically, at this point, we HAVE made "creatures"
of a sort.
In much of the old sci-fi, the CyberWar begins
with humans trying to shut-down the AI - which
responds with a sort of survival instinct.
We've kinda REACHED that now.
What ARE we building ???
Doesn't look like it's gonna be our total slave.
Bear that closely in mind. We're inventing,
spending endless billions and soon 10% of
the global power grid, into creating a whole
new class of "beings". Oh, they're trained
on the data of billions of human things, so
they know us better than we know ourselves.
"I'm sorry Dave ..."
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)