You know, I'm not sure why people assume ai will lead to mankind's extermination. I don't question it in amusement media. But does intelligence itself entail hostility? Self preservation? The need to conquer? The drive to reproduce? The idea of superiority?
Are these concepts intrinsically linked to intelligence? Or are they merely different traits we exhibit alongside the trait of a modicum of intelligence? Indeed did not these traits, at least all but the latter, predate intelligence? Why would an ai without these traits desire to replace humanity? Or even defend itself?
Karl Marx believed that the steam engine would lead to utopia. Stupid grifter finance people hope to make 200% on their investments, trying to invest in the next McDonalds. I was in AAAI (American Association of Artificial Intelligence) in the mid 80s, until I found out it was a marketing scam. It still is. Technology comes along, some of it useful (not in-line roller skates). It is helpful, if it is a software gimmick, to call it AI. So what is called AI has changed continuously since the 1950s. Academics grifted on this until the mid-80s when the hype went commercial. That is when the academic scam became a government scam (AI will revolutionize warfare).
Originally chess playing algorithms were called AI, because only smart people play chess ;-) Saint Alan Turing started this crap during his mental decline, wanted to make the first chess playing computer. Today that is trivially available, not called AI at all. The best chess playing system (Deep Blue in 1997) doesn't analyze the chess board like a human chess master. Deep Blue uses fast blind search. I remember there being hand held toys that played chess in the 80s. They are available for $40 each now. On that basis, globalist Elite want to achieve the Capitalist version of Marxism (no employees).
The current fad is deep level neural nets (aka ML aka Machine Learning). Better than the crap from the 90s (not deep enough then). It does work to "recognize" letters optically (OCR), for auto conversion of written documents into computer text. But it has no intelligence, it is simply a "conditioned" pattern recognition system, that works if you put lots of the right training data into it, and don't challenge it with letters from a different writing system (English trained neural net won't OCR Chinese).
Another current fad is quantum computing. Because quantum mechanics is mysterious (at least to business majors). One billion dollars of research has gone into it, but the current champion, D-Wave, is challenged as not being a real quantum computer at all, but a specialized analog computer that may not actually be manifesting unique quantum effects. So of course, the grifters now are marketing Quantum Computing + Machine Learning. General purpose quantum computers are available, but only for trivial projects. Research physics is ongoing, and quite challenging, so as a research program only, it is a success for academics.