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.