There are cycles of AI summers and AI winters in the biz. In between the advances that yield "summers" there are pullbacks and periods of lack of investment in between ("winters" ).
First there was direct programming. It was able to find a few things people had not, with mathematical theorems or simple robots.
Then came expert systems for things like medical diagnosis. They have not gone away, but of course have their limitations.
Now we are at LLMs. While there is hype and there is discounting of "hallucinations", the systems are consequential and applicable to more contexts than just chat or search. People are expecting too much from them and there will inevitably be some disappointment which is already beginning to show up.
An AI winter is coming when disappointment becomes dominant, but it does not mean AI will go away or AI is not very powerful. LLMs applied to other domains than language are already finding new drugs, new alloys, new device designs. Those are found but then they are vetted by humans who analyze and run trials or run mechanical tests.
AI, in its current form, is very real though over-hyped. Anyone who writes it off is making a big mistake.
= new reply since forum marked as read