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.
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.
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.
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.
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"...
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.
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.
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.
Safari? Of all browsers? Why not start with Chromium?
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.
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.
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"...
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...