Can AI art be recognized? Does it have a place on this site?

What about this tiny bed leg? It’s the only one. They must have extremely uneven floor there.

3 Likes


Don’t forget this mirror that defies the laws of physics (and what is this lamp doing?)

4 Likes

Hmm, I’m not so sure, but what I’ve seen is that people will often make a big deal out of something, time will pass, that thing isn’t as big of a deal or it has some major limitations of some kind (because everything has limits), and then even if things are a problem, people figure them out eventually.

3 Likes

No more tequilas for the AI. Who said to put the carpet upside down?

It’s a carpet, isn’t it? It’s hard to tell.

4 Likes

Let’s hope it can’t walk away on its own with all that weight on it.

1 Like

Notice also that the bed isn’t touching the floor at the head, it’s at a 20ish degree angle and floating

2 Likes

No fair, I’m still on that last image! This is a fun game.

Where I think we need an AI is in text analysis so I don’t keep getting moderated for using entirely INNOCENT words! I’m starting a list.

3 Likes

Which one is the worst error of all?

1 Like

If you had any doubt about whether ripped game models are allowed on this site, just try replying to this and using the word “ripped” :wink:

1 Like

The worst error is calling any of these “art”

2 Likes

You know what I think will be funny? Someone “working hard” on creating those images and looking at at them and thinking: “That’s really good!” then coming to forums like BA and showing this as if they could fool anyone. I would just laugh.

2 Likes

Yeah, those images would do fine on Twitter but I’m fairly confident in this site’s ability to spot AI images :slight_smile:

2 Likes

Thank you for confirming my guess. :wink:

1 Like

I have perhaps an unfair amount of power. Watch this: ripped NFT nsfw naked nude…

I could go on but I don’t want to get in trouble :sweat_smile:

3 Likes

I think it’s really hard for the AI to deal with human creativity and interpret what we create. It’s significantly easier to understand what a human looks like. But when it comes to a chair, it’s not so easy. We create chairs in so many shapes and different numbers of legs, I imagine it’s hard for it to understand where it can use and how it can make a chair with just one leg or 6 or whatever. Also about the rules of ergonomy. It only creates images that at first glance look ok.

1 Like

@piranha4D Is writing for about 15 minutes. I imagine the newspaper that’s coming. :joy::joy::joy::joy::joy:

23 minutes later… Still writing. :joy::joy::joy::joy::joy:

1 Like

:snrk: Funny, and I deserve that I suppose. A long phone call rudely interrupted my composition. :wink:

Sort of, except it doesn’t interpret. If the AI has a large enough dataset, with precise enough text descriptions, then it can form the sort of latent space containing information that defines “chair-ness” in many ways, plus latent spaces for other words you use in addition, from which it can derive any kind of chair-facsimile you might desire. If you actually ask for a “chair” in your prompt, you’ll likely get one that’s better than the thing in this image. But if you specify a superset of which chairs are just a part, it might come up with some unholy combinations. Because it doesn’t really understand. It does a mathematical analysis.

Which is, in part, what makes it such a fun concept creation tool, because it doesn’t necessarily derive concepts from the same narrowed down categories we often do. It has a huge space filled with object-ness descriptions from which it can draw conclusions, and those are sometimes… innovative. But it doesn’t do that intelligently. It does that according to algorithms it has been fed. That’s why I don’t really like the term AI. Machine learning is a long way from actual intelligence.

The more highly curated the dataset, the better the AI can do a synthesis, but usually those datasets are relatively small, and the AI can only do very specific things. You could make one that’s excellent at chairs – but it couldn’t do anything else, not even sofas.

It’s actually pretty hard to define a chair well enough to cover all eventualities. I wonder in how far those huge dataset models fail because they scrape public information and that just isn’t the best way to teach (it’s cheap though, and probably the only way to teach an AI lot all at once). But maybe I just don’t know enough (I definitely don’t know enough).

4 Likes

I’m glad I waited for the answer. :wink: I didn’t know it was for me, though.

1 Like

Topic removed due to directed bullying.

Topic removed due to directed bullying.