People were posting low-quality rambling bullshit, sometimes completely off-base, without even bothering the most basic of smell-tests. People occasionally post low-quality rambling bullshit too, or things that are off-base, but with ChatGPT you can post 100 answers in an hour.
It's a matter of scale. The ban wasn't pre-emptive, it was reactive in response to a real observed problem with people lazily Ctrl+C/Ctrl+V spamming poor quality nonsense from ChatGPT.
While technically not allowed, you can still use ChatGPT on Stack Overflow: just make sure it's correct, copy-edit things a bit to remove some of the waffling and repetition that ChatGPT tends to generate, and no one will even notice.
I'm less involved in the art community, but I would imagine that most communities are at least in part about people who create things for the joy of creating things, and then share that in the community. I don't have anything against AI art, but if lots of people start lazily spamming that kind of stuff then you've kind of lost your community. It's not so much about what is or isn't art, it's about having a community.
That said, this mod is clearly being an ass about it.
The biggest problem with ChatGPT is that when it's wrong, it's confidently wrong and cannot quantify its uncertainty in any way (maybe it's too human-like in that respect...) Furthermore, the whole idea of a reputation economy collapses if reputation becomes "too cheap to meter".
I wish 90% of real people answers were correct...
I see AI-generated art as being similar to taking performance enhancing drugs in sport or using something that's against the rules in motorsports. Outsiders don't really care because they just see someone performing at the same level of the others by using clever tech, but if you're part of the group then you will care much more.
As an unpopular opinion: I don't watch races, but maybe I would watch that. I mean, Roborace is pretty cool (but it's mostly about the tech rather than racing, and IIRC it's all strictly driverless), I wonder how a man-machine pair would do.
Similarly, I find most sports boring and hardly worth watching (unless in a good company, but then it's not really about the sports), but if some day there would be Olympic Games NG+ without any doping or gear restrictions, I'd most likely check those out with great interest. I do realize there would be tons of drawbacks and nuances involved (and more than a lot of people screaming hate how this is wrong to them), but that'd be a) interesting on a personal level - watching some folks who trained for their whole life isn't really resonating with me - I'm happy for them being beyond good but I can't say I really care; and b) would produce enormous effects on society, in terms of medical and technological advances, just like the space races of the past (and hopefully the future) - I'd surely cheer for that.
Then, I'm a proponent of machine assistance in computer games, too. In my opinion, human bodies and minds are inherent sources of unfair advantage and machine assistance - if equally available to everyone - is the greatest equalizer. Though, of course, I acknowledge that a lot of games are designed solely or mostly around imperfections of human performance (mechanical or perception).
I'm a software dev attempting to learn art. I recently joined Mastodon related to this purpose and it's quite the hot topic there. Many, many artists pissed about how their work is being used to train corporate profits as well as potentially undermining their living/passion/etc. I've actually seen some cool art in protest of "AI"... usually involving malformed hands which the artist community have gravitated towards being the representation of current AI capabilities.
I think it matters how it is created, personally. Not because the author of an individual piece of art is important to me, but rather because once AI moves into a problem space and can effectively and accurately "solve" that problem space the displacement of humans will be surreal. How it affects people is the important thing to me. I'll be interested to see how we manage to recognize this reality as AI improves.
"Who cares if the diagnosis is done by a medical expert or someone pulling out random drugs they tried before? As long as I feel better immediately after taking them, who cares how they were prescribed?"
The case on SO is clearly different, as ChatGPT might answer incorrectly or answer with something containing subtle bugs. There's also a good chance that you won't be immediately able to spot those bugs, as, if you were sufficiently knowledgeable in the topic yourself, you would have most likely not asked that question.
The case for art is a bit different, as there is no technically correct way to do it, but there is still a value to the way it is created. Would you think the first picture drawn by your child is worth the same as any other bad painting? Would you agree that a perfect copy of the Mona Lisa has equal value to the actual object? If no, it should be pretty easy to see why a painting generated by an AI is different from one created by a human.
Playing devil's advocate here, but this also clearly applies to human answers.
As a test I ran some questions through ChatGPT a few weeks ago; both popular ones and just some random ones from the homepage (including some not-very-good questions). In only one case would I say that it was "good enough" to post. All the rest ranged from "mostly correct but with huge omissions" to "used 5 paragraphs to explain what could be one or two sentences" to "this is not even remotely correct in any way, but at a glance it actually looks kinda correct". The "looks kinda correct" can be pretty misleading, because there have been a few answers where I went "wait, is this actually a thing? I didn't know about that!" and when investigating further it turned out it wasn't a thing at all and ChatGPT was just trolling me.
- Humans actually try to answer the question at hand, whereas ChatGPT is only auto-completing text. This might lead to similar results in a lot of cases, but the goal of ChatGPT is not to produce correct SO answers.
- Related, humans will (usually) back off or correct answers if errors are pointed out to them. ChatGPT (at the moment) doesn't.
- Humans will (usually) not attempt to answer questions they don't feel like they are qualified to answer. ChatGPT will.
I'm not saying that SO is/was perfect as-is, but ChatGPT is - in my opinion - overall not likely to improve SOs quality as of now. Since it is currently banned on SO, it seems like this is not only my opinion.
If you submit an answer yourself and it’s wrong, if someone begins the process of critiquing it or editing it, they can engage in a dialogue with you in order to make this happen. You can explain how you came up with your answer, and they can help you debug your thinking. Seeing this process unfold over a couple comments is often one of the most enlightening things on SO.
How is this supposed to happen if you submit a ChatGPT answer which you have just accepted on blind faith and maybe don’t even understand?
I don't think the average person cares at all.
It's the same with banning ChatGPT from StackOverflow: Who cares and who notices? Art is either evoking some feeling or not and it's different for everybody. An answer on SO is either helpful or not. Who cares how it was written? ChatGPT can easily say something more helpful than me, stable diffusion can easily make something I'd rather have on my wall than Da Vinci's Mona Lisa (or anything more along my preferences). Why do we care so much? What's "real art" anyway?
I always like a colleague's mousepad, it said: "Is this art or can we throw this out?" Always makes me smile.