It’s not that LLM will do that exact thing, they have a big dataset so they know how “eleven” works. If you trained a NN with this small dataset, than this will happen or rather it’s a comment on how AIs work in general. With more complex things something like this might happen and then you have this meme to relate to