where it draws from infromation and it's own experiences and makes something new,

It does not. The main thing is that it just does not, because it can't. There is nothing inside that could do that.

Like chatbots you could interact with, and they reply or AI that writes things. The AI would not be able to do those things if it was just simply programmed to carry out specific actions.

Check old toy program named Eliza written in Perl (based on 1966 concept) long ago. Today it is available as perl module Chatbot::Eliza in CPAN. It is pretty simple thing, less than few screens of code. It did exactly same thing in mid 90s around internet without any hype. I had run it on forums to make luls from users who had long discussions and arguing with it without having clue that they talk with a program. It is even more advanced than modern AI, because it builds structure of concepts and you are able to teach it the meaning of new things and it will use them in dialogues.

Both, Eliza and modern ANNs are exactly programmed to carry out specific actions, no more no less.

You miss the point completely. As soon as most people will be convinced in a lie that not-an-AI have some more intellect than light switch, all fucking bastards who rule the world will immidiately transfer all their responsibility to that not-an-AI to avoid any even highly improbable consequences. That is the only purpose of all that false AI narrative.

All research in areas of computer science that could really move us close to creation of something that could really be named "artificial intelligence" was curtailed in mid-90s and ANNs become the only allowed mainstream. There was many promising things in that area, you could even heard about something like "expert systems" or "problem solvers" - that was the traces of that lost technology. Today, you will find nothing except dumb and fully controlled ANNs that could not produce anything new in principle, only repeat things that was put into it by humans.