I notice that the token in question happens to be segmented as “_an” and “_a” and not “_an_” or “_a_”.
So continuations like [“_a”,”moral”,”_fruit”] or [“_an”,”tagonist”,”ic”,”_monster”,”s”] could be possible (assuming those are all legal tokens).
I am reminded of the wonderful little nuggest in linguistics, where people are supposed to have said something like “a narange” (because that kind of fruit came from the spanish province of “naranja”). The details on these claims are often not well documented.
Something that might be tried is to look for neurons that are very likely to be active two or three steps before a given token is actually produced, which might represent a sort of “intent for the token to show up eventually”.
If you find such things, I might suggest that they be named “Wernicke Neurons”, and I wouldn’t be surprised if they ended up being highly/visibly related to “an intelligible semantic intent” in sketches of active speaking personas who probably want to eventually/potentially touch on >1 keywords.
The thing I would expect is that many related keywords would have prioritization before they are uttered, and they would hold each other somewhat in abeyance, with jostling and deal making amongst themselves, maybe with some equivalent to Broca’s Area as the judge of who should win when various Wernicke Neurons try to make their ideas “go out the door” first (or at all)?
I notice that the token in question happens to be segmented as “_an” and “_a” and not “_an_” or “_a_”.
So continuations like [“_a”,”moral”,”_fruit”] or [“_an”,”tagonist”,”ic”,”_monster”,”s”] could be possible (assuming those are all legal tokens).
I am reminded of the wonderful little nuggest in linguistics, where people are supposed to have said something like “a narange” (because that kind of fruit came from the spanish province of “naranja”). The details on these claims are often not well documented.
Relatively more scholarly analysis of such “junctural resegmentation” issues does exist though!
Something that might be tried is to look for neurons that are very likely to be active two or three steps before a given token is actually produced, which might represent a sort of “intent for the token to show up eventually”.
If you find such things, I might suggest that they be named “Wernicke Neurons”, and I wouldn’t be surprised if they ended up being highly/visibly related to “an intelligible semantic intent” in sketches of active speaking personas who probably want to eventually/potentially touch on >1 keywords.
The thing I would expect is that many related keywords would have prioritization before they are uttered, and they would hold each other somewhat in abeyance, with jostling and deal making amongst themselves, maybe with some equivalent to Broca’s Area as the judge of who should win when various Wernicke Neurons try to make their ideas “go out the door” first (or at all)?