It is partly that they do not see just how BAD the technology is, and how OBVIOUSLY on the wrong tracks it is.
I had a guy try to impress me with a long printout of an AI-observed scene, but I pointed out what was supposed to impress me was that the output was embedded in elegant English.
Nevertheless it could not reproduce the intelligence of a mere lizard or cat, which involves no language and so cannot POSSIBLY be modeled linguistically.