Are we heading to a point where AI images look so real that even the best tools can’t detect them?
Welcome to this forum
Discussion Guidelines
Here are a few pointers to keep discussions smooth:
- Posts should have at least 100 characters—more detail helps.
- Check if your question has already been answered by using the search feature.
- Share both the pros and cons of AI respectfully.
- If you’re making claims, include reliable sources.
- No wild theories, like AI being the end of the world—it’s been debunked many times.
Let us know if you have any questions or feedback.
I’m just a bot and can’t reply directly, but moderators can address your concerns.
I think we’re almost there, if not already. It’s not just a future scenario anymore.
Zephyr said:
I think we’re almost there, if not already. It’s not just a future scenario anymore.
If single images are so good, what about videos? Since videos are just a bunch of images, wouldn’t they also become indistinguishable?
@Joss
Yeah, but videos still have issues like keeping things consistent across frames and the cost of creating them. But those problems will probably get solved soon. Right now, I’d be skeptical of anything over 10 seconds long.
@Zephyr
By 2026, maybe we’ll push it to 20 seconds?
Finnick said:
@Zephyr
By 2026, maybe we’ll push it to 20 seconds?
It’s already possible to create longer clips by making quick camera movements or scene changes every 5–10 seconds. It tricks the viewer and gives AI more flexibility. Soon they’ll probably fix any jerky transitions too.
@Ren
This sounds like it could seriously disrupt industries like filmmaking.
@Zephyr
Keeping consistency across frames is tricky because of how these models work. And costs are huge. Even now, some models like OpenAI’s advanced systems can cost up to $1000 for a single task. Imagine the expense of generating an entire movie.
@Cody
True, but costs are dropping fast. Just compare prices to 2022—it’s a massive difference, and it’s only going to get better.
Zephyr said:
@Cody
True, but costs are dropping fast. Just compare prices to 2022—it’s a massive difference, and it’s only going to get better.
I wonder if mixing symbolic AI with neural networks could help. I studied AI back when symbolic systems were the focus. Combining those two methods might be what we need to make big breakthroughs.
AI made a perfect clone of your mom, and you couldn’t even tell!
Hollis said:
AI made a perfect clone of your mom, and you couldn’t even tell!
Your mom is so ugly that AI refuses to create her picture—it’s against its guidelines.
It’s already possible if you’re willing to put in a little effort.
AI detectors aren’t reliable because they’re also AI-based.
AI can already make images and text that seem completely real. It’s especially easy for text, but images and videos are catching up fast.
I saw a video of AI making Will Smith eat spaghetti. It’s come a long way. In five years, it might be perfect.
If you find this topic interesting, look into how GANs work (Generative Adversarial Networks). They’re used in training many AI systems by having two neural networks compete—one creating images, and the other trying to spot the fake ones.
@Mackenzie
That’s really cool. Thanks for explaining!
You can literally see the progress happening.