As someone who works in IT, I agree.
Pattern recognition algorithms and wordings arranged according to syntactic rules but with no comprehension of the meaning.
There is no intelligence in an algorithm.
The big danger is if companies use "AI" output as if it were fact and without need for scrutiny.
If you're going to scan the internet for the input then there's a lot of misinformation there.
If you're not going to validate the training data and feedback loop then the system will likely reinforce incorrect content and conclusions.
A recent example is Apple's fake BBC news story.