[NitpickLawyer]

> No matter the current state, boosters believe it will get better

How can it not, tho? Like it can't get worse (at least for the open models. You have them now. The underlying models don't change). So the only obvious answer is they either stagnate, or get better. And a brief look at history shows that nothing truly stagnates.

Keep in mind that we're not even 3 years into this new paradigm. We're still scratching the surface of how we can use these things, that are unreasonably good for being trained for NTP.

That's my main argument against an "AI winter". We have so many things to try that it'll take decades to fully realise all the potential of this tech, even if all research into foundational models would stop now. And it surely won't. There is just too much capital allocated here, and that attracts a lot of smart people working the problem. Yeah, it's reasonable to say things will get better.

[techpineapple]

“And a brief look at history shows that nothing truly stagnates.”

This seems so false to me we must be looking at different histories. Almost everything stagnates, we just over index in the things that don’t.