Re: Whoopee.
Bubbles inevitably burst. I get the sense that the "AI" bubble doesn't have long left, given that most of the use cases seem to be pretty illusory when you look at them closely.
Tech companies have invested heavily in the "free sample" model, but are going to have to start charging soon, and people just aren't going to pay to get ChatGPT to summarise emails for them, or for Bing image generator to come up with yet another variation on "Man of Pider".
This isn't to say that it has zero uses, but the cost/benefit analysis doesn't weigh up for most of them. I used Copilot yesterday to generate a simple test harness for an API client I've been working on. It needed quite a lot of finessing just to get it to compile, and even then, the code is a mess, and likely full of inefficiencies and vulnerabilities (ironically, Copilot also pointed several of these out when I put in the PR on GitHub). Given that it's just something internal to assist our QA testing, because the suppliers of the actual API are dogshit to work with, this doesn't matter, but it neatly illustrates the limited usefulness of AI, even in the few cases where it is actually useful. It may have saved me half a day or so in dev time, but does this weigh up against the monthly cost my employer is paying for it? Will it continue to do so when the inevitable price hikes come along.
Meanwhile, I'll put off rebuilding my gaming PC with a new motherboard and RAM until prices come back down to a sensible level, in the firm knowledge that RAM prices have always gone up and down like a rollercoaster. The 5070ti I've got my eye on will no doubt be a few hundred quid cheaper in the new year anyway.