We hear a lot from two camps on AI: the "boosters" and the "antis".
From those two camps we get a sense that AI is going to go in one of two directions. The boosters say it'll exponentially get better at everything until human intelligence is meaningless and we we suddenly have to work out how to do a UBI. The antis say that it's all going to collapse in a flaming heap when we all finally admit it can't do anything and when the bills for those datacentres come due.
Either way, the outcome is extreme - a world full of AI, or a world empty of it.
I've been wondering if there's a third, worse option: that zombie AI just keeps ambling along for years, maybe a decade, until we all look back and admit "well this is all a bit crap, isn't it".
I mean, look at what we mostly get from AI now: crappy software delivered at pace, crappy writing that we're all developing allergies to, crappy fake pictures and videos jamming up the Internet and training us to feel cynicism instead of wonder, crappy product reviews and how-to pages that swamp the real ones and mean that the Internet is rapidly becoming useless, crappy chatbots instead of humans when we have a problem with our utility bills.
So what if instead of getting way better, or way worse, it just... stays this way? Never getting good enough to fix things, but just barely adequate enough that no-one is prepared to throw it away.
The tech sector keeps pushing the crappy software because it keeps engineers' bargaining power in check, and hitting your velocity metrics is so much simpler than doing the hard work of actually making something good. Social media doesn't crack down on AI because it's the only thing left that might increase engagement. The education industry pushes it because hey, it might be a bit worse, but it's a lot cheaper - and have you seen how uppity human teachers can be? Social services get overtaken by it, because what politician has ever run on paying more for public servants?
I think part of the problem is that a lot of industries have incentives to use AI, even if it's worse. Sure, they might be hoping it gets better, but they're not relying on that - and won't stop using it if it doesn't.
So I can pretty easily picture us getting to the end of the decade and thinking "wasn't this supposed to start getting better at some point?"
But by that point it might be too late. A generation of software engineers don't know how to code, and there's no structure to teach them. Who even knows who you hire to design a curriculum anymore? You have a problem with your driver's license but you just ignore it because the chatbots go in circles when you ask about it, and no-one knows how to contact a human. You have a lump in your armpit, but the medical AI won't let you see a human doctor - there aren't many left - until you've done a few more rounds of blood tests.
Companies were handed the "make it a bit worse, but a lot cheaper" button and many of them started pressing it as fast as they could. We're all hoping that AI at least gets good enough before things go too far, but what if it doesn't? Will we turn around? Or just accept that things kind of suck now? And how much might we lose along the way?