This is great. Having the option to choose whether or not to see AI content was the first thing I hoped they'd add a filter for when it was first introduced.
My only concern is that I have seen legit videos by creators on the platform for many years, with the tiny "AI (i)" label below the video. I have seen this even on channels that are not faceless.
You would think if anyone could get this right, it would be Google/YouTube, but apparently not.
My pet peeve is the AI generated voices. I absolutely love the cosmology-astrophysics-space stuff. And it can be any crap if it's correct enough and entertaining enough and passive enough that I can fall asleep to it.
But once I detect that the voiceover is AI generated that shit get instablocked forever. It's an instant put-off. Do all the visuals you want with AI as long as they are reasonable but when a human (AI) voice starts to fuck with the way a word has been presented in a particular context then it just makes me want to hate everything I won't mention any names <cough> elevenlabs.
This is great, but why start with Safari and not a cross platform browser? Is there something native here that can't be done otherwise or is it just personal preference?
There's plenty of AI blocking Chrome extensions already, however I prefer Safari for personal stuff (and hence YouTube too) and use Brave for Google-gated work related stuff.
This is what they mean when they say AI amplifies inequality. If you're technically competent, you just write yourself an extension to block the slop spam, or at least pay to install one. So in the end it's rotting much more the brains of the least prepared.
I’ve found that the AI label doesn’t necessarily mean AI slop, it just means there’s a non zero amount of AI in the video. I want to avoid AI slop but not all videos that have an AI title screen or whatever.
Yeah, I make song remixes in the style of a particular video game and while the music is completely human written (by me), I'll often have AI generated or edit the "cover art" for me so people have something to look at/chuckle at.
These videos are still marked as AI, and fairly so; however I don't think that means people should want them hidden :-(
Google’s been pushing creators to use AI thumbnails more heavily since right around the EU deadline for labeling, presumably in the hopes of saturating the label so that everyone burns out and eats their slop quietly.
Yep, it's super useless. The label is so general. Was AI used for one image in the 10 minute video? Is the voiceover AI? Is the entirety of it generated with AI? No way to know.
Agreed. I think instead of "AI or not", that something more nuanced is needed - perhaps a rating system inspired by the MPAA's (g, pg, pg-13 r, nc-17, x) that is used for movies.
For a creator posting a video, Google gives specific guidelines on whether to tag it as AI:
"Was AI used to generate or edit your content in any of the following ways?
- Makes a real person appear to say or do something they didn’t say or do
- Alters footage of a real event or place
- Generates a realistic-looking scene that didn’t actually occur
To follow YouTube’s policy, you’re required to tell us about your use of AI. This includes realistic sounds or visuals made or edited with AI. [View examples](https://support.google.com/youtube/answer/14328491)"
So oddly an intro graphic only has to be tagged as AI if it shows a real/realistic person, place, or event, but (per the examples page) apparently any AI-generated music counts.
True, but that's a way harder problem to solve. For me what was important was reducing the noise as much as possible as fast as possible consistently. Youtube algorithms has been pervasive enough that anything that moves the scale down even a little bit was worth the effort.
It's free on GitHub for anyone to build, use or fork. I've built it for myself, but going through App Store submissions took me 2x the time, hence the price.
one problem is that algorithms(1) to detect AI generated content are surprisingly bad
like non-AI generated animated/cartoon content is much more likely to be mistakenly labeled as AI then some somewhat realistic "looking" but for a human still clearly AI slop content
some platforms (TickTock) seem to label anything as fully AI generated iff the metadata indicates certain tools like Adobe Illustrator are used, independent on how it was been used
---
(1): If that tool is "ask an AI" then its not surprising at all. It's content which AI tried to make look as realistic as possible, the same (and potentially similar) AI will see it as "as realistic as possible"...
- stupid question: how does this sort of thing work?
- like how do you actually determine if a video was AI made or not?
- what is the algorithm used? what is the threshold for a video to be considered slop?
- what do you do? download the videos with yt-dlp split them into frames and run each frame by what?
This is great. Having the option to choose whether or not to see AI content was the first thing I hoped they'd add a filter for when it was first introduced.
Just for kicks, I inverted the filter in a TamperMoneky script that only keeps the AI-generated content: https://gist.github.com/SMarioMan/32c959d79035cb421df1c69c06...
My only concern is that I have seen legit videos by creators on the platform for many years, with the tiny "AI (i)" label below the video. I have seen this even on channels that are not faceless.
You would think if anyone could get this right, it would be Google/YouTube, but apparently not.
I’d love a small local model that detects AI voices and blocks the content.
My pet peeve is the AI generated voices. I absolutely love the cosmology-astrophysics-space stuff. And it can be any crap if it's correct enough and entertaining enough and passive enough that I can fall asleep to it.
But once I detect that the voiceover is AI generated that shit get instablocked forever. It's an instant put-off. Do all the visuals you want with AI as long as they are reasonable but when a human (AI) voice starts to fuck with the way a word has been presented in a particular context then it just makes me want to hate everything I won't mention any names <cough> elevenlabs.
I wish I had this but for every website and media surface
Well, AI flags this content so it's an easy target without heuristics - that's why I've started there.
Ironically, if you had this for every website, including GitHub and HN, you wouldn't have seen this HN submission / repository :)
That's also true, you gotta fight fire with fire :)
Need the EU, or California or some other large jurisdiction to create a law requiring identifying labels on all AI content
This is great, but why start with Safari and not a cross platform browser? Is there something native here that can't be done otherwise or is it just personal preference?
There's plenty of AI blocking Chrome extensions already, however I prefer Safari for personal stuff (and hence YouTube too) and use Brave for Google-gated work related stuff.
This is what they mean when they say AI amplifies inequality. If you're technically competent, you just write yourself an extension to block the slop spam, or at least pay to install one. So in the end it's rotting much more the brains of the least prepared.
I’ve found that the AI label doesn’t necessarily mean AI slop, it just means there’s a non zero amount of AI in the video. I want to avoid AI slop but not all videos that have an AI title screen or whatever.
Yeah, I make song remixes in the style of a particular video game and while the music is completely human written (by me), I'll often have AI generated or edit the "cover art" for me so people have something to look at/chuckle at.
These videos are still marked as AI, and fairly so; however I don't think that means people should want them hidden :-(
Google’s been pushing creators to use AI thumbnails more heavily since right around the EU deadline for labeling, presumably in the hopes of saturating the label so that everyone burns out and eats their slop quietly.
Yep, it's super useless. The label is so general. Was AI used for one image in the 10 minute video? Is the voiceover AI? Is the entirety of it generated with AI? No way to know.
Agreed. I think instead of "AI or not", that something more nuanced is needed - perhaps a rating system inspired by the MPAA's (g, pg, pg-13 r, nc-17, x) that is used for movies.
Now that the AI label exists, the creators using AI for an intro graphic or background music are probably going to stop using it.
For a creator posting a video, Google gives specific guidelines on whether to tag it as AI:
"Was AI used to generate or edit your content in any of the following ways?
- Makes a real person appear to say or do something they didn’t say or do
- Alters footage of a real event or place
- Generates a realistic-looking scene that didn’t actually occur
To follow YouTube’s policy, you’re required to tell us about your use of AI. This includes realistic sounds or visuals made or edited with AI. [View examples](https://support.google.com/youtube/answer/14328491)"
So oddly an intro graphic only has to be tagged as AI if it shows a real/realistic person, place, or event, but (per the examples page) apparently any AI-generated music counts.
I'm more concerned about the stuff that is fraudulent then possibly legit stuff explicitly labelled as AI
True, but that's a way harder problem to solve. For me what was important was reducing the noise as much as possible as fast as possible consistently. Youtube algorithms has been pervasive enough that anything that moves the scale down even a little bit was worth the effort.
You can charge money for browser extensions? And people pay for them?
I'm in the wrong business.
It's free on GitHub for anyone to build, use or fork. I've built it for myself, but going through App Store submissions took me 2x the time, hence the price.
They require signing which is $100/year.
https://support.apple.com/guide/security/app-code-signing-pr...
one problem is that algorithms(1) to detect AI generated content are surprisingly bad
like non-AI generated animated/cartoon content is much more likely to be mistakenly labeled as AI then some somewhat realistic "looking" but for a human still clearly AI slop content
some platforms (TickTock) seem to label anything as fully AI generated iff the metadata indicates certain tools like Adobe Illustrator are used, independent on how it was been used
---
(1): If that tool is "ask an AI" then its not surprising at all. It's content which AI tried to make look as realistic as possible, the same (and potentially similar) AI will see it as "as realistic as possible"...
Safari? Of all browsers? Why not start with Chromium?