[LordDragonfang]

The huge difference between this and your analogy is that 3d printing failed to take off because it never reached mass adoption, and stayed in the "fiddly and expensive" stage. GPT models have already seen adoption in nearly every product your average consumer uses, in some cases heedless of whether it even makes sense in that context. Windows has it built in. Nearly everyone I know (under the age of 40) has used at least one product downstream of OpenAI, and more often than not a handful of them.

That said, yeah it's mostly niche locations like customer support chatbots, because the killer app is "app-to-user interface that's undisguisable from normal human interaction". But you're underestimating just how much of the labor force are effectively just an interface between a customer and some app (like a POS). "Magical" is exactly the requirement to replace people like that.

[cogman10]

> But you're underestimating just how much of the labor force are effectively just an interface between a customer and some app

That's the sleight of hand LLM advocates are playing right now.

"Imagine how many people are just putting data into computers! We could replace them all!"

Yet LLMs aren't "just putting data into a computer" They aren't even really user/app interfaces. They are a magic box you can give directives to and get (generally correct, but not always) answers from.

Go ahead, ask your LLM "Create an excel document with the last 30 days of the high temperatures for blank". What happens? Did it create that excel document? Why not?

LLMs don't bridge the user/app gap. They bridge the user/knowledge gap, sometimes sort of.

[j2kun]

"Adoption" of tech companies pushing it on you is very different from "adoption" in terms of the average person using it in a meaningful way and liking it.