Asklemmy

There are tells but it's getting harder and harder. One thing is, you have to look for the people.

Case in point, a few weeks ago I discovered some really good covers of the KPop Demon Hunter songs. So if you don't know the scene, this Netflix musical has been making record breaking profits and numbers and everyone and their brother wants a piece. And the music is so good, and tons of people are doing covers.

A few months ago, "is it on Apple Music/Spotify" wouldn't have been a tell, but now it is. So a lot of these covers are on paid streaming, because covers are perfectly acceptable legally. It's fair use. However, recently AI generated music has started to come up, and the streaming services have recently put their feet down and said "no more." So when you're looking at such a cover on YouTube, it's probably going to have streaming links. It's more convenient to listen on one of those services than it is to watch YouTube, and Spotify pays more than YouTube, and Apple Music pays more than both of them. So that's where they want you to listen. When people are all over the comments saying "put it on Spotify" and they say "nah we're YouTube exclusive," what they're saying is, they aren't allowed on Spotify (or Apple Music). They do want more money, but they won't get any on platforms that ban generative AI. So they stick to YouTube, which is one platform that allows it.

With legitimate art, you can usually find the human behind it. With some art, the artist will want to remain anonymous. In anime, for example, a lot of these artists are underage, and they're savvy, they're not putting themselves out there on social media beyond the art and their anonymous comments. They're still more human than AI, they just won't show their face or say where they're posting from. (And that's just good OPSEC in general.)

There's also the frequency. Art takes time. It takes weeks, months, depending on what it is. AI can do it in seconds. So if they're posting whole new stuff every day, every other day, there's a solid chance they're using AI to make it.

Not all AI slop is completely made by AI. Sometimes they take stuff made by humans and use AI to enhance it somehow. That's what the KPDH stuff was. They were using an AI tool to separate the stems (the individual instruments) and enhancing each one, changing some, altering others.

Anyway, in 2025 now, it's much harder than it ever was before to spot AI slop. The time of six fingered hands is gone. Next year, year after, certainly the next decade, it's gonna be next to impossible to tell.

What's worse, these days, AI made stuff is still prompted by humans. All AI slop has a human behind them. But what happens when the AI starts doing this stuff on its own? Right now, AI is interacted with by humans. Soon, AI will initiate the interactions. It could be doing it already, we don't know.

Form meaningful trusting relationships with other people

Long story short don't let an algorithm choose the content you watch.

If you are holding the book in your hand and it was published before 2015, then that content is almost certainly the product of human action.

I recently made a pdf with some of my notes on hints of AI in images and music, but I'm not sure how to send files here

It's not easy ofc, and it will get harder with time, but I am convinced we can tell if trained a bit. Because there are clear differences in the creative processes between humans and machines, which will always result in different biases

With time I think we'll learn to only trust people we have some social connection with, so we know they are real and they don't use AI (or they use it up to a level acceptable to us)

Don't watch short form content. Try not to watch faceless youtube channels unless they are obviously real and trusted.

In the future you won't be able to tell. This is about the time you'll have to worry about the fate of humanity.

https://vger.to/lemmy.world/c/slopornot

Computer algorithms have been enhancing movies and pictures for a while now.

Also: could some interesting Lemme create /c/slopornot ?

I'd like to preface that I like AI content for my own amusement and sometimes for convenience. I think it's neither the best nor the worst thing to ever happen to the world like so many lemmy users seem to.

Current state, most AI generated content (images, video, even text) have some general tells. For text, they tend to lean on certain phrases and formatting. Picture and video both still contain noticeable artifacts that give it away, though that is becoming less prevalent over time. They're a lot more noticeable when you use the tools yourself and trying to overcome the patterns is difficult without manually intervening.

I think you have to ask yourself what degree of human involvement is the cutoff for you. Is it only 100% non generated content? Even prior to the sudden llm push, that would be really difficult to find. A lot of software, photoshop and predictive text for example, have used machine learning to improve their algorithms for years. It's not likely you'll find anything unassisted anymore. What degree of human made modifications AFTER something is generated is enough to consider it good enough? If I start with a generated image but significantly modify it with an image editor to fix issues and finalize the 'vision', is that enough?

I personally think you'll have to create your own compass. Bad content is bad regardless of how it's made. If you cannot tell the difference if a human made it or a machine, does it really matter?

The tools used in creation have next to nothing to do with the quality of the output.