AI does not give people a blanket performance bump - situationally it can be used in ways to enhance productivity in some tasks, I've had good success using it to chart out things like big database moves and some of the VM parts of MVVM builds - but that time savings is cut down heavily by having to hand check the results, and for many complex or especially novel tasks (very common with database/analysis tasks, not very common with MVVM tasks) they're absolute money-burning dogshit.
I think the pushback you're getting here is due to you being kinda abrasive about it, and glossing over the areas wherein AI is the aforementioned absolute money-burning dogshit. We'll undoubtedly be keeping AI in some ways - we can run it locally, there's open source models, etc. but this tech is getting fairly mature at this point, and we haven't found any new novel areas in which it's useful. It's shit for data analysis, every implementation of general agentic AI has been nothing but an expensive quagmire, image generation will never be profitable outside of simple models (and of course porn) but AI is situationally really quite useful as a code assistant, for infosec, for NLP/chatbots/image recognition and it will 100% be sticking around (in some forms) for all those use cases.
It's just not a black and white issue like many pro-AI folks here have been presenting it, and you've been deeply patronizing in your explanation of your perspective on the issue. Hence, pushback.