[A4ET8a8uTh0_v2]

<< it isn't yet a << yet

That is a lot pressure to put on a conjunction. It is up there along with 'it will never be'.

In all seriousness ( and some disclosure ), I like this tech so I am mildly biased in my stance. That said, I almost fully disagree with yours.

As much as I dislike Nadella, his last blog entry is not that far off. Using LLMs for stuff like email summaries is.. kinda silly at best. The right use cases may have not emerged yet, but, in a very real sense, it already has been transformative..

[hunterpayne]

"it already has been transformative.."

Yea, at being a search interface. But what else? Not that it can't be, but the failure rate for AI is absurd right now. What happens if it collapses and all its used for is answering questions on your phone and maybe better search of your emails? That seems to be a real and probably likely outcome. What then? Ironically, I think it will improve the economy because there are a lot of decisions that are on hold until we know what LLMs will be used for. Probably isn't going to be good for SEs either way.

[A4ET8a8uTh0_v2]

<< but the failure rate for AI is absurd right now.

I keep a personal log of specific failures for simple CYA reasons. I do get some, but I can't honestly say it does not seem high to me. A lot likely depends on what is defined as a failure ( to me it typically is a clearly wrong result ). But those clearly wrong results do not seem to cross 10% of output.. so about the same as average human.