Gary Marcus alludes to clever hans in an article discussing Ai. I think this is THE example with Ai. Clever hans was a horse who could apparently solve maths questions. But, it was tapping it’s hooves according to onlookers’ reactions.
That’s precisely what current models do. Look around, sense reactions and give results without understanding the context at all.
Clever hans could in theory perform all maths calculations which needed a whole real integer answer where the value was say less than 100. Ai is in the same spot. It just does not understand what a horse is.