I am all for supporting local artists and I feel that “handcrafted in XXX” products make great souvenirs when you’re connected to those places. Still, if some AI hallucinated me a perfect novel for my interests, or generated something I couldnt tell was manufactured or created by a master, I would happily enjoy it.
“How can I tell if this is slop so I can know to hate it” sounds stupid to me: good is good. When it comes to art / food / products, I want the best experience for ME. If I want human connectedness, then I’ll go interact with a human directly.
I can do without wasted water, power, and money, but in the abstract it seems to bother everyone on Lemmy to enjoy something a person didn’t make. I don’t have that hang-up.
In the abstract, I don’t really disagree with you. It definitely wouldn’t replace human made things or human perspectives, though. (which, as a human myself, I value highly). But I do agree that If I could have a genie show me the most beautiful picture possible according to my tastes, I would still be very interested in seeing it even if nobody made it.
This is a complete hypothetical though. The problem is that the generative models we have now aren’t actually good at much of anything, beyond purely functional stuff like boilerplate text or code. (and even that is somewhat debatable)
Firstly, nothing these models produce is art, which requires a perspective that they do not have. It isn’t valuable aesthetically or or in a literary sense, either; what is produced will always be bland and unchallenging. To oversimplify, it is the ‘average’ of the training data - nothing original or innovative.
I think one big issue that isn’t talked about enough is that it can look okay at first, but is actually shit. It ends up being a kind of information poison. It makes it harder to parse information, because useless junk is no longer obvious. Like finding what looks like a fresh apple, but is actually just the skin with nothing inside. Especially with llms, there are so many ways it can be subtly bad.
What scares me the most is it improving in the future to the point where I can’t tell and it’s poisoning me without my knowledge. It’s noticeable already - I’ve listened to podcast interviews where the guest has picked up on chatgpt mannerisms, and it’s really eerie, because it’s destroyed their personal voice and they don’t even notice. “AI psychosis” is definitely a real thing, and I think it’s also affecting a lot of people more subtly.
While this is a bit of an exaggeration, I see why you dismiss this at first. However, it’s someone recognizing that they may not be able to sort high quality information from low (or not wanting the burden of doing so). Knowing it came from a human not only means it has some inherent value, but also means that we will be proficient in judging and sorting our own informational input.
This is a lot of text to not say anything and do it out of both sides of your mouth.
There literally was, in the last 4 days, a post in no stupid questions asking what to do when you can’t tell a “slop post” from one written by a person. I wanted to make the premise of this post a comment and I wish I had so I could link it.
There is plenty of text related stuff that any LLM does better than most people on Lemmy. To include writing comments, essays, and probably poetry to be honest because people are ABSOLUTELY TERRIBLE at meter and counting syllables. Boilerplate code is also good, but the state of LLM related output in response to GOOD promoting is pretty high. Of course if you ask how many rocks to eat in a day or ask for something in a dumb way it’s garbage in, garbage out.
All of that notwithstanding, if there is something decent that comes from an LLM or generative AI that doesn’t mean it must be flawed or that it needs to be hated. Every generated piece of text is not flawed nor does it contain inherent poison. That’s the whole point.
My issue is that when I read, I’m doing it to update my perspective on things. Web comments, even if they are poorly considered or written, contain information about how people think or feel about stuff; for instance your comment here references a post you saw recently and uses it to give more context to your argument. But say your comment had been AI generated, and done well enough that I can’t tell the difference - now I believe that post existed, and that someone had this reaction to it, but in this scenario that would have been all made up, and what I believe has been distorted a little further away from reality, and my judgment has gotten a little bit worse.
When I see articles here that appear to be written by AI, it doesn’t help if it meets heuristic standards of writing quality, what matters is that its references are fake and the context of its arguments are kept deliberately vague to mask how it is devoid of salient information. Having well organized rhetoric just makes this take longer to figure out, misinformation isn’t made better by being more convincing. I agree with you that people can be a little too dogmatic about rejecting anything with any sort of connection to AI, but it isn’t as simple as, if you enjoy it it must be good.
I think we can both agree that AI methods have no place in original journalism without oversight and input from actual journalists.
That said, for articles in general, not every source is fake if it comes from a language model. That’s the whole point of Resource Augmented Generation (RAG) for example.
While you may glean some perspective from reading comments here, I would say that 1) not every perspective is that good or worth assimilating, and 2) even though as far as you know I’m a person, there’s still no reason you can trust that the post I referenced actually existed. It’s hard to say that misinformation is made worse or better by AI, people who want to lead you somewhere wrong will still do it.
Yes there are. I’ve seen plenty of posts like that, so it’s plausible. It could be verified without too much effort, so even though I’m not going to go to the effort, it would be a risky lie to tell, a different lie would make much more sense. There’s not a lot of reasons for you to write all this other than being actually frustrated with people’s sentiments on this issue, and considering that the apparent strength of an argument isn’t likely to change many people’s minds here anyway, the emotional logic for that to be a deception doesn’t really check out.
I would say that the big difference is the depth of misinformation, especially in the context of informal writing. Someone can have an agenda and try to work out how to manipulate people in the direction of their agenda. Someone can want to fuck with people and tell arbitrary lies for their amusement or feelings of dominance. But in all those cases, what they say is still a product of who they are, what they are thinking and how they are feeling, and that information comes through in one way or another. For all of that subtle information about the state of people to be buried in noise all coming from the same source is a terrible thing.