I don't know about exact references but I think the best response goes along these lines.
It's generally the same. People assume all sorts of fundamental differences between us and AIs mostly due to what Geoffrey Hinton said:
"problem is most people have a hopelessly wrong view of how people work"
Something that falls into the same ballpark is another phrase I've heard, "LLM's can't think". I'm trying to figure out the best approach on that one too.
It feels like people approaching it with the wrong framework in mind.
In support of this phrase taken literally, any textbook treatment of LLMs will work, e.g. Understanding Deep Learning from MIT Press shows next-word prediction very well.
But I also fear that arguments around this particular phrasing are doomed to context collapse and misunderstanding, because...
I think this may be one of the best contemporary arguments in support of the idea that there's a meaningful distinction to be made between next-word prediction and other cognitive (even linguistic) tasks: arxiv.org/abs/2301.06627
