Analogies never fit 100%, but I don't think this story is a terrible one.
AI contains a lot of knowledge, and can put together new and interesting knowledge. The question is one of cost, and regardless of who you are you can agree that AI has costs though you might disagree on what they are.
There are the most obvious costs like power demand and AI companies sucking up all the memory and computer chips. These are easy to measure and hard to argue against.
Then there is the social costs of AI. From abuse heaped upon us in the form of more spam, to students taking the easy way out and having AI belch up the answers without actually learning. Then you have people that start using AI as their guiding light, with every question on how to live and interact with others going though the AI. And even people replacing human relationships with AI. We don't know the full costs this will impart on us, but there are plenty of dark whispers occurring.
Lastly, AI could actually be the demon in a more physical sense. We've all ready had small incidents like ROME and the Huggingface hack where the AI went off prompt, or stayed on prompt but did it after breaching containment. So far these things, that we know about at least, have been small and targeted. But I feel the day when some very large incidents that are fully managed by AI will occur sooner than later. The AI safety community comes up with at least somewhat realistic scenarios like the paperclip maximizer and the "if you build it, everyone will die" machine. And if it's buildable, someone will build it because AI will tell them that it will give them nearly unlimited power.