AIs like ChatGPT are extremely expensive toys. OpenAI receives hundreds of millions of dollars in donations every year. Yet, the GPT family of AIs are only useful for generating text that is of dubious utility/accuracy. This is especially obvious when generating code, because it always requires a programmer to fix it. At best ChatGPT can cut down on the time to generate text, but a human will have to clean it up.
But ChatGPT can't do anything else. It can't prepare your taxes (it doesn't have uptodate tax codes). It can't diagnose what disease you have (it can't look at you and see your symptoms). It can't plan a delivery route. All ChatGPT does is randomly generate text that resembles text seen on the internet. To a lot of people, ChatpGPT looks like magic -- often it's inexplicably funny. But it is only a glorified chatbot. As such, ChatGPT (aka GPT-4), like it's predecessors GPT-1 through 3, is a technological dead-end -- a flash in the pan that will be replaced by another flash in the pan called GPT-5.
But there are other forms of AI that are not toys -- there are AIs that are very useful:
  • Google Search is an AI that has stood the test of time as not just useful, but it has revolutionized how we use the internet.
  • Expert systems are quietly used in various industries, especially in diagnosing diseases. Expert systems have been around for decades and have saved humanity unknown amounts of money, yet few find them interesting, because they aren't fun like a toy.