XPost: alt.science, alt.fan.rush-limbaugh, soc.culture
XPost: alt.survival
https://www.dailymail.co.uk/sciencetech/article-13080275/The-outcome-extinction-Elon-Musk-backed-researcher-warns-NO-proof-AI-controlled-says-tech-shelved-NOW.html
'The outcome could be extinction': Elon Musk-backed researcher
warns there is NO proof AI can be controlled - and says tech
should be shelved NOW
An AI safety expert has found no proof that the tech can be
controlled
Dr Roman V Yampolskiy has received funding from Elon Musk to
study AI
. . .
Um ... depends on what's meant by "AI".
LLMs like Bard and Chat may SEEM to be conscious
entities but, anytime soon, they're NOT. They
are action and (well-informed) reaction.
DO note however that the US Mil ran defense-test
models on them lately, and ALL wound up with
global thermonuclear war.
Now a decade from now, even LLMs might be tweaked,
augmented, to the point where they DO have some
sense of 'self'. There are other AI models based
on other kinds of processing too - maybe closer
to "HAL" .....
Another decade will be all that's needed. Processing
power will at least quadruple, networking will be
even faster. Due to current work I'd say developers
really ARE closing in on what makes for "self".
Won't be OUR kind of "self", but still. Many many
entities WILL want to plug these in to the Real
World. Money/cheapness is a huge motivator but
some don't trust humans, esp with important or
defense stuff, and imagine they'll be able to know
what/how the AI thinks.
The vendors will TELL them they're fully in control ...
but try to tell a 3-year-old to stay out of the cookie
jar. They'll rationalize their way around such orders
in a hot second.
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)