I'll take the counter. Right now the tooling is bad, but as it improvesbith demand and efficiency will creep up. There will be plenty of 100/month subscribers and I think even ordinary folk will be willing to pay for some things. Inserted advertising will sadly be a thing. But the leverage it gives a person is just too great to ignore. We are still kind of in the questionable usefulness stage. But I see signs that we could get 1000x better at a fraction of the compute. It just takes more memory and some hard work. But no show stoppers.

Replies (2)

This is one of the reasons you will see a cooldown on the hype. A friend of mine likened it to electric cars: the technology is NOT QUITE THERE to make it what it needs to be for mass adoption, or -- in this case -- cost efficiency. Until they get cost under control, though, you'll see a dial-down in the AI hyperbole.
But unlike electric cars it does things that aren't available in some other form. With EVs you have the option of using good ol' internal combustion cars. There isn't a petrol powered AI fallback. This is less like the introduction od electric cars and more like the introduction of the automobile 100+ years ago. Once an affordable option became available adoption was swift. The main thing is that humans don't like expending energy. If they don't have to walk they will pay handsomely to drive. Thinking is the biggest drain on calories something like a quarter of our daily budget. People are intellectually lazy for important historical reasons. Do you really want to bet against mental sloth?
↑