Research shows people rate AI-generated content as better than human content, until they find out it's AI-generated. : futurology

[–] AceFuzzLord@lemm.ee 7 points 2 years ago (2 children)

Based on my use of AI chatbots for story writing (chatgpt not included for different reasons), I personally have a somewhat negative view in general of the fact that, to me at least, the stories are average quality. Another downside is they pretty much all end with a happy little moral like message like "Benny found from that day forward he would work extra hard to make sure to do the right thing after he realized his mistakes." Almost always, except with my early tests with chatgpt, ended with anywhere between a sentence or a short paragraph with a happy little message.

Also, I am also one of those people who is less likely to like your art if it looks similar to an AI generated image/style because I absolutely do not like how soulless it feels compared to art an actual artist has made.

[–] kromem@lemmy.world 3 points 2 years ago

It's because you are using a chat/instruct fine tuned version of the LLM.

Don't think that represents what is actually currently possible for creative writing with a SotA LLM.

Part of their fine tuning step was literally to make it come across as more soulless than it was in the pre-release version, which was still a chat/instruct model (just an earlier version) and yet coming across as too human-like to users.

[–] Sylver@lemmy.world 6 points 2 years ago* (last edited 2 years ago) (4 children)

A big part of my appreciation towards the arts is the creativity behind it. Obviously I am not impressed with an internet-searching AI when compared, unless I am specifically interested in the code of the AI

[–] CanadaPlus@futurology.today 9 points 2 years ago (1 children)

How much do you know about the code of AIs? There seems to be an idea that they just photoshop together stuff they find off the internet, which is nothing like the actual situation.

[–] Sylver@lemmy.world 2 points 2 years ago (2 children)

As a CompSci major I at least have a great idea of what’s under the hood, so I can appreciate the creator of the AI and those that helped push the field to its current limits.

[–] FaceDeer@kbin.social 6 points 2 years ago (2 children)

Well, you should know in that case that AI art generators don't "search the internet" when they generate images. I have a local installation of Stable Diffusion and it works perfectly fine with the internet disconnected.

[–] dumdum666@kbin.social 3 points 2 years ago* (last edited 2 years ago) (1 children)

To be fair: to create said local model on your pc it had to get trained on thousands if not millions of accurately described images. You only install what the model has learned on your computer- not the training data.

[–] FaceDeer@kbin.social 3 points 2 years ago (1 children)

Sure, but the person I'm responding to literally described AI art generators as "internet-searching AI" and that's a common misconception of how they work. Even among people who confidently claim to know how these things work.

[–] dumdum666@kbin.social 0 points 2 years ago (1 children)

Yeah - didn’t read his first comment… CompSci Major…

Sorry…

[–] FaceDeer@kbin.social 1 points 2 years ago

It's okay, there are a lot of misconceptions out there and what you said was also informative and true. So hopefully it all helps.

[–] Sylver@lemmy.world 1 points 2 years ago

It appears I did a lot of damage by over-generalizing in my comment. My bad.

Either way, that Stable Diffusion was trained on shared data, not that it literally requires Googling and scanning pictures.

I wasn’t trying to make it a lesson on AI so my bad! Thanks for the info!

[–] Daxtron2@startrek.website 3 points 2 years ago

Comp sci majors are perfect examples of the dunning Kruger effect when it comes to anything tech related.

[–] WraithGear@lemmy.world 5 points 2 years ago* (last edited 2 years ago)

I judge art more by how it affects me. But then again i do separate beauty / something that i can find meaning in from art by an intention. And AI has no intention, or an intention to create at least. A spider web covered in morning dew is beautiful and can contain meaning but its not art.

[–] cm0002@lemmy.world 5 points 2 years ago

Art is incredibly subjective and comes in many forms, I consider AI art to be just another form of art. And as a person who can barely draw a stick figure, a form of art that I can actually use to get ideas out of my head for once lmao

[–] FaceDeer@kbin.social 2 points 2 years ago

How do you judge how much "creativity" is behind any given image? You say you're "obviously" not impressed with AI-generated art, but how is it so obvious to you which images are AI generated and which are not? There are plenty of quizzes online that challenge people to identify AI art, this one for example, and it certainly doesn't seem obvious to me.

[–] dumdum666@kbin.social 5 points 2 years ago

AI created stuff is sometimes quite good and tops low level writing in pretty much every case … I personally think that there are many ways we can profit from this culturally.

We have to fight all those greedy assholes though, that think they can now put wages under even more pressure because of AI.

[–] CanadaPlus@futurology.today 2 points 2 years ago (2 children)

Hah! Of course we do. It's like how people love cheap swill poured into an expensive wine bottle.

[–] FaceDeer@kbin.social 9 points 2 years ago

An apt comparison given that "wine tasting" is bunk.

When it comes right down to it, subjective experience for inherently subjective things like taste should be all that matters. Million-dollar price tags for art or bottles of wine are purely performative, things having high value simply because everyone has collectively agreed that it should have high value (often for purposes of tax evasion or social standing).

If there's a bottle of wine or a piece of art that I enjoy the experience of, what does it matter how it was created?

[–] DavidGarcia@feddit.nl 1 points 2 years ago (1 children)

so you're saying human made art is cheap swill?

[–] CanadaPlus@futurology.today 1 points 2 years ago* (last edited 2 years ago) (1 children)

I mean, the study is kind of saying that. I don't have a particular reason to doubt the result, since I've seen some pretty spectacular AI content.

Edit: On further thought, maybe it's more that human content is a mediocre wine, and AI content is a similar but maybe slightly better wine, but "human" is the expensive label.

[–] DavidGarcia@feddit.nl 1 points 2 years ago

it's definitely better than the average "artist", even the average artist, but the top artists are still leagues above. Anyone saying otherwise is lying to themselves, as evidenced by this study.

What I hate most is that this discussion is so black and white. If any amount of AI is involved it's allegedly garbage. You can create 99% of an artpiece on your own, add some detail with AI and it's automatically garbage.

Makes no sense. People will just have to accept that the nature of making art has changed forever and realize that it's not about the process, but about the end result. And if you like doing art 100% manually, no one is stopping you.

You average person won't be able to create good art even with AI, as evidenced by the AI art communities on Lemmy. Being an artist means having an eye for beauty. You can't teach that.

That's one of the main reasons why people think AI art is garbage, because it's mainly created by people with skill or zero taste. In the hands of real artist, AI art turns out amazing. I wish I had anywhere near that aesthetic gift.

Any idiot like me can tell you if art is good or bad, but creating good art is an entirely different beast. It's extremly rare, even among professional artists.

That still requires the human touch. AI can't come anywhere close.

[–] BilboBargains@lemmy.world 2 points 2 years ago* (last edited 2 years ago) (1 children)

How have we arrived at the situation where machines are making art while people are toiling in shit warehouse jobs? Shouldn't the machines be working for us while we relax and make art?

[–] WarlordSdocy@lemmy.world 1 points 2 years ago (1 children)

I mean it makes total sense when you look how much wages are in those two areas. Creative fields tend to pay well when you're successful at it while the warehouse jobs are low paying. Makes sense for a corporation to want to get rid of those higher paying jobs if possible.

[–] BilboBargains@lemmy.world 1 points 2 years ago

There is some logic to that argument but it rarely applies to the top management. In my field, ten thousand engineers will make some cars but ten thousand managers will make zero cars. It's something I regularly remind them of.

We've been told there is something special about our management since the dawn of time. It used to be warlords, barons, kings and queens. We were told they were backed by divine authority. Meanwhile, people like Shakespeare were creating the real stuff of value and we don't know what he even looked like.

[–] Wilzax@lemmy.world 1 points 2 years ago

AI is better but it's made by cheating. No good model was trained only on content that was public domain or properly licensed for use, so the images it produces are effectively a blend of traced artworks and never an original work.

[–] Heavybell@lemmy.world -3 points 2 years ago* (last edited 2 years ago) (1 children)

Yeah this doesn't shock me. Generative AI is gonna be trained on the best art possible, so of course you're gonna get good looking output… until you realise the thing that created it doesn't actually understand 3D space, or find other imperfections that reveal it for the thorough cargo-copy it is.

[–] kromem@lemmy.world 1 points 2 years ago (1 children)

u tol you realise the thing that created it doesn't actually understand 3D space

You might find the following paper interesting as the reality is a fair bit more nuanced than you might think:

Language Models Represent Space and Time

[–] Heavybell@lemmy.world 2 points 2 years ago (1 children)

Will give that a read in the morning, thanks. I am only talking about the generated art I've seen, which often features a clear lack of understanding of 3D space. When I see generated art that shows understanding, I'll be impressed.

[–] kromem@lemmy.world 1 points 2 years ago (1 children)

Ah, you mean diffusion models (which are different from transformer models for text).

There's recent advances in that as well - you might not have seen Stability's preview announcement of their offering here, and there's big players like Nvidia and dedicated startups focused on it as well. Expect that application of the tech to move quickly in the next 18 months.

[–] Heavybell@lemmy.world 1 points 2 years ago (1 children)

Yeah, I didn't thin LLMs did art generation.

[–] kromem@lemmy.world 1 points 2 years ago (1 children)

Actually, the transformer approach was just used with some neat success for efficient 3D model generation:

https://nihalsid.github.io/mesh-gpt/

[–] Heavybell@lemmy.world 2 points 2 years ago

That looks cool :)

Futurology