It's an overpriced, expensive novelty toy at the moment.
Some of the smaller models are useful for a little things. But most of those run on CPU and aren't really all that helpful for any real.
And two or three generations of gpus and improvements to hardware models that are currently considered frontier probably will run on reasonable home equipment. Not your Grandad's computer, but like a high-end gaming PC or workstation more than likely.
At that point we'll probably hit the good enough threshold where it'll be useful and reliable enough to do mundane tasks with little to no error.
And that's probably where all this is going to end up after the bubble popping. It's all just going to kind of be mundane and boring. It's going to change the world but not in any overly large way at least not quickly.
It's probably just going to be like the.com bubble and everything is just going to slowly shift over like a 20-year span.
The whole business model of a giant company selling AI subscription honestly is cannibalizing itself when you step back and look at it. They're making more efficient models that are more compact and faster generation after generation at a ridiculous speed.
The moment a new generation of hardware comes out to run these things even faster, you've suddenly just massively hurt your ability to sell the lower end of your stock.
Given enough generations of hardware and models. You'll reach a point where the small models are doing what the giant models could do generations ago. At speeds faster than those models could do generations ago.
And what the hell does your average small business even need with a super fast super big model when something they can run off computer that they have right there that they own in their own office can just do.
You just end up at a point where you can't sell your product. People aren't going to need frontier class models forever. The better the mid-tier wants to get and the better the low tier ones get the less of a value proposition the frontier ones become.