Wikipedia does not appear to describe the term AGI as being used only by a, ah, clique or uh, relatively small faction, or however you would describe “ Ben Goertzel et al, and the LessWrong/MIRI community”.
Perhaps this is because the editors of the Wikipedia article on “Artificial General Intelligence” are aligned with LW-ers.
However, the article seems well cited, and says that ‘the term “artificial general intelligence” was used as early as 1997, by Mark Gubrud’ (and provides a citation for this).
Ah, ok, Wikipedia also says the term was re-introduced and popularized by Dhane Legg and Ben Goertzel, around 2002. (Sorry if this is nit-picky, I’m writing this as I read about the topic, with only a little editing of parts I’ve already written.)
Wikipedia in particular defines the term as referring to the general capabilities idea, and its article on AGI doesn’t even seem to mention the specific architecture that Goertzel proposed involving ‘a network whose nodes and links carry probabilistic truth values as well as “attention values” ‘ (to quote the article which is about Goertzel himself) .
So, Wikipedia at least gives the impression that there is a fairly clear consensus on at least generally what “AGI” means (even if not on the exact details of a definition / criterion. It lists a number of proposed tests which seem similar enough in terms of the kind of things they test.)