https://futurism.com/the-byte/government-ai-worse-summarizing
The upshot: these AI summaries were so bad that the assessors agreed that using them could require more work down the line, because of the amount of fact-checking they require. If that's the case, then the purported upsides of using the technology — cost-cutting and time-saving — are seriously called into question.
To expand on that for people who think it's all just smoke and mirrors. I think, just like the assembly line, work places will be reorganized to facilitate the usefulness/capabilities of LLM's and, perhaps more importantly, designed to obviate their weaknesses.
It's just that people are still figuring out what that new organization will look like. There hasn't been a Henry Ford type for LLM's yet (and hopefully won't be a Nazi this time). Obviously there's no guarantee there will be such a person/organization but I don't think it super unlikely either.
Yeah the hype cycle is certainly annoying. As is the accompanying fire/re-hire at lower pay cycle that follows any automation.
I actually think the generative aspect of neural networks is the least interesting/useful/innovative/etc. Though it will admittedly be more interesting when an LLM can say, use blender to make a video rather than just wholesale generating it. Or at least generate the files/3d models necessary to have it be edited by a person just like they would anything else. I suspect there will have to be a pretty significant architecture change for them to be able to make convincing/coherent movie-length videos.
Chaotic system control, like they're doing with nuclear fusion plasma is the most interesting, to me anyway.