Hacker News Reader

Comment by jmward01 | Hacker News Reader

I hate to admit it but I failed the NPR real vs fake video quiz [1] and it is exactly because of this. There is so much fake noise out there that it is very hard to tell what is true.

[1] https://www.npr.org/2025/11/30/nx-s1-5610951/fake-ai-videos-...

View on HN

Thanks for sharing this. I got all 4, but none of which were so obvious that I had absolutely no doubt. I had to reason about all of them. And I'm absolutely confident that a LOT of perfectly reasonable people can potentially score zero on this test.

Same. Managed all 4. But the differences are tiny and I'm only 70% confident. Most of my judgement is based on human reactions to a changing situation.

Yeah, I have pretty much stopped analyzing the media itself for cues, and am evaluating the scene and the actors. Are they convincing? The behaviour of the cops in the first video were entirely unconvincing. I didn't consider the video quality, artifacts, lip sync issues, etc.

Thanks for sharing. I am curious which of the four quiz questions you failed—to me they looked relatively easy to tell apart, but I follow the progression of this tech very closely.

Personally, I've mislabeled the one with the animal in the restaurant as AI generated. I might have clicked too quickly because it was looking like the animals trampolines video's. I've not really looked at the timestamps.

I'm generally good at detecting AI generated content but I might have a few false positives. :)

4/4 for me, but probably only because I knew about that tear gas incident, and that snake was way too loud and they don't tend to move like that unless it's sand.

The tear gas was the only one that had me guessing. Knew the video was real, but wasn't sure it wasn't doctored just at the end with the throw. Overall, it read more real than fake, I was just sure they were going to try to "gotcha" me.

Reply on news.ycombinator.com