Subj : AI a Lie
To   : OGG
From : Rob Mccart
Date : Fri Oct 10 2025 08:54:54

RM> ...was given limited time for whatever it was working on at the
RM> moment..  Things seemed a bit off a while later and they checked
RM> and found that when the system couldn't complete a job it was
RM> working on in the allotted time, it actualy rewrote part of its
RM> own programming to give itself more time to complete the work.

OG>Nah..  it didn't rewrite anything. It's totally possible that
  >the original code failed to implement a hard fast rule when
  >time runs out thus allowing the process to continue.

I decided to have another look for that original story because
I figured if it was a common problem then it wouldn't have been
newsworthy enough to get on a National News broadcast..

So, this may be a case of both of us being right at some level
but the original story involved Sakana AI and it mentioned the
potential risks related to AI autonomy when their AI 'Attempted'
to modify its own code to extend the runtime of it's experiments,
which they said could lead to unexpected behaviors and challenges
in control.. and they go on to needing more robust safety
protocols and isolating AI systems from critical infrastructure
to prevent unintended consequences.

So the battle begins.. Every time people try to rein in what
their AI can do on its own, it will try to find a way around
that..  The old better mouse trap = better mouse problem..  B)

---
 þ SLMR Rob  þ click...click...click..Damn... out of taglines!
 þ Synchronet þ CAPCITY2 * capcity2.synchro.net * Telnet/SSH:2022/Rlogin/HTTP

.