It's painful to see otherwise very smart people still stuck on #1. Talk of sticking your head in the sand.
@woozle "inappropriate usage" is a legitimate concern for almost every piece of technology & software though so I don't understand why that's presented as if it only applies to LLMs.
Also the "creative theft" argument is IMO quite silly, given that the whole Free Software movement was based on the philosophy of "you can't own an idea". What **is** beyond unfair is that copyrights and IP matter when they benefit the big companies but that's nothing new and nothing actually specific to AI.
As for the pollution issue, sure but notice the huge intersection between the anti-AI and anti-nuclear energy crowds. It sometimes feels like these people just don't like technological progress of any kind. Our energy demands are expected to rise with technological progress: https://en.wikipedia.org/wiki/Kardashev_scale
@zeenix When it comes to "creative theft", I feel like that's an argument for UBI.
The Agreeing Part
As someone who has been creating music, art, and writing for many decades now, and only recently getting paid even tiny amounts (relative to a living wage) for any of it, I'm not bothered at all by the idea that some of my stuff might end up going into the mix for producing new stuff. I mean, that's how music and art have always worked.
If LLM output can be shown to be too imitative of any individual piece of art, then that is a problem -- but as far as I can tell, people are unhappy even about output that is clearly a synthesis of many other works.
The Disagreeing Part
On the other end of things: from what I can tell, the level of datacenter expansion is truly unnecessary to provide even the level of LLM services currently available. It reflects a bug in capitalism, rather than technological or infrastructural progress. They seem to have reversed all gains towards carbon reduction, when we could have required them to be carbon-neutral (or even carbon-negative).
TLDR: It's not the level of energy-consumption that is the problem; it's how that energy is being produced.
I frankly find it alarming when I read about natgas or coal plants being reopened or even freshly built just to power datacenters -- not to mention the way the owners seem to care not at all about how much water they evaporate for cooling or the negative impacts to adjacent communities.
This issue feels very real to me -- but it's not an indictment of the tech itself; its an indictment of the techcorps, and of capitalism in general.
"inappropriate usage" is a legitimate concern for almost every piece of technology & software though so I don't understand why that's presented as if it only applies to LLMs.
It's legitimate to point out the specific ways in which it applies to LLMs -- or, more precisely, the ways in which LLMs are being used inappropriately.
...though I agree that this again is not an indictment of the technology itself, but more a valid warning. "Do not hammer people or delicate objects unless you intend to break them." "Do not use fire without understanding the dangers and how to prevent them." ...and so on.
@zeenix I think... that beyond the legitimate concerns about datacenter pollution, creative theft, cognitive surrender, and inappropriate usage, there is also an at least quasi-legitimate philosophical anxiety about what it would mean if LLMs are actually an advancement towards genuine artificial sapience. ...and that anxiety is like an emotional magnet that pulls people towards any argument which seems to disprove that possibility.