The explosion in data centres around the world is effectively undoing all of the modest steps towards reducing carbon dioxide emissions. We're probably already in an unavoidable death spiral for our way of life but continuing down the path we're on so that people can have a computer write their emails for them and make pictures of naked celebrities means that we'll probably see that collapse in living standards a matter of decades.

This sub: LLMs don't work for the benefit of humanity and are a waste of energy. Also this sub: the extinction risks from irresponsible AI ~~research~~ deployment are intolerable.

FTFY.

"Why don't the various members of this fedi community all think the same thing? Checkmate!"

There is an extinction risk because LLMs don’t work. They are next word predictors, so putting them in charge of things that can kill people is a real problem.

(To be clear, LLMs have some real world use cases, but they don’t work when given too much authority because they don’t understand the consequences of what they’re doing. So “don’t work” is entirely contextual.)

Your conment assumes both that a. critics agree with the ai boosters that an extiction risk is posed by the technology and b. the nature of that risk is a high level of ai functionality.

The most realistic doomsday scenario is a large-scale version of the incident recently where AI hallucinations nearly led to the U.S. aatacking a Chinese ship. LLMs are kinda trash to mid at most use cases they're but towards, that's part of the risk they pose.

The point of this isn't even whether or not the risks are real, the point is critiquing the psychopaths building shit that they believe might kill us all. AI boosters' brains are literally too fucking cooked to understand the critique.