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.
Hey there, I've got no stakes here and I don't want to speak for anyone but I think what happened here was QuillCrestFalconer and DPRK_Chopra were simply pointing out that the technology is rapidly evolving, that it's capabilities even just a couple years ago were way less than now, and it appears that it will continue to develop like this. So their point would be that we need to still prepare and anticipate that it may soon advance to the point where employers will be more willing to try to replace real workers with it. I don't think they were implying that this would be a good thing, or that it would be a smart or savvy move, just that it's a possible and maybe even a likely outcome. We've already seen various industries attempt to start doing that with the limited abilities of "AI" already so to me it does seem reasonable to expect them to want to do that more as it gets better. Okay, thanks for reading. 👋
Yeah, the technology is rapidly developing but I am not the only one unconvinced that just piling in more data in the exact same way as it is now is going to 1:1 match biological brains. I'm not saying it is impossible, far from it. I'm saying the current "just spend more energy and produce more carbon waste pile on the data" approach, powered by marketing, isn't likely to produce a generalized artificial intelligence on its own.
Marketing hype being what it is and how it's both misused and even doing a disservice to actual nascent artificial intelligence research, I reject calling the current LLM technology "AI."
Okay. I am under no illusion that current technology is anywhere near replicating digital brains. I don't think that's what QuillcrestFalconer or DPRK_Chopra were saying either. When we say "replace workers" we mean "replace the functions that those workers do for their employers". We're not talking about making a copy of your coworker Bob, but making a program that does many of the tasks that are currently assigned to Bob in a manner that isn't too much worse than the real guy (from the warped perspective of management and shareholders of course), and anything the machine can't do can be delegated to someone else who gets paid a pittance. That's what we're talking about, nothing about recreating human intellects. I put the term AI in scare quotes in my first comment because I too am well aware that it's a misnomer. But it's the term that everyone knows this technology by (via marketing and such like you said) so it's easy fall back on that term. LLM, or "AI" in scare quotes, I don't think the specific term really matters in this context because we're not talking about true intelligence, but automation of task work that currently is done by paid human employees.
My primary beef and the main thrust of my argument was exactly that: the primary triumph of "AI" is as a marketing term.
It does a disservice to research and development of generalized artificial intelligence (which I hope won't be such a fucking massive waste of resources and such a massive producer of additional carbon waste and other pollution) by jumping the gun and prematurely declaring that "AI" is already here.
I think it does, unfortunately, if only because of how people already take that misleading label and ride it hard.
Valid discussion for sure, and I wish it could be pried away from the marketing bullshit because it's really misleading a lot of people, including otherwise educated people that should know better.