AI videos that are animated, unrealistic, or only have a little AI may still hide their origins, though.
See full article...
See full article...
YouTube to begin automatically labeling AI videos
- Thread starter JournalBot
- Start date
t_negative1
Ars Centurion
ok,
but when its wrong. i'm sure its going to be
a hassle to undo it.
also, what about the videos with CGI that arent AI,
i'm sure they will get flagged too.
none of this matters to me. but there are plenty of
people that care.
but when its wrong. i'm sure its going to be
a hassle to undo it.
also, what about the videos with CGI that arent AI,
i'm sure they will get flagged too.
none of this matters to me. but there are plenty of
people that care.
Upvote
3
(15
/
-12)
Call me picky, but PLEASE, folks who did this video, get someone who doesn't do a CONSTANT AND CONSISTENT wave, wave, wave, hand-hold, wave, wave, wave, hand-hold, wave, wave, wave hand-hold motion while talking though the entire thing.
Give him a fidget spinner or something. I literally did not hear a word he was saying the first time through because his presence and gestures were so distracting.
More topically, I guess this is fine, but my standard approach to anything on the Internet is that it has to meet a minimal burden of proof and Google's assurances don't meet that standard.
As long as they have any financial incentive to fudge things in any way, the bar they have to clear can't even be reached by SpaceX. This, of course, assumes they decide to keep it instead of fucking with the soup 297129873 times before they finally toss it and say, "you're on your own!"
Upvote
-17
(16
/
-33)
jonbob_newcastle
Ars Scholae Palatinae
But of course they won’t be using this technology to weed out deepfakes and AI slop when it’s used for scam ads.
Upvote
75
(75
/
0)
You know how ages ago, web browsers would display an adorable lock icon to indicate the page you were on used SSL (Secure Something Layer) encryption? These days, the majority of the WWW (World Wide Web) utilise HTTPS (Hypertext Transfer/Transport) Protocol Something) so the larger browsers don't flag the ubiquitous SSL sites anymore by default.
How long until GenAI is so ubiquitous that they simply tag what's likely NOT GenAI?
Upvote
26
(26
/
0)
This does seem like a step in the right direction.
I would also like them to stop or add a toggle to remove the horrible AI filter/sharpening whatever that they run over some videos. Make them look terrible and this is for videos from channels I know are legit and aren't using AI.
I would also like them to stop or add a toggle to remove the horrible AI filter/sharpening whatever that they run over some videos. Make them look terrible and this is for videos from channels I know are legit and aren't using AI.
Upvote
18
(18
/
0)
It's a start. Another nice thing would be to add 'don't recommend this channel' to the options while you're actually watching a video. That's my go-to for AI channels; black hole them.
Unfortunately I'm not always able to detect that before starting one, and once I've done that it's usually difficult to find them to do it.
Unfortunately I'm not always able to detect that before starting one, and once I've done that it's usually difficult to find them to do it.
Upvote
35
(35
/
0)
I am. I saw a video from a creator who learned a few phrases in Spanish for a video/joke. She uploaded it, and Youtube automatically auto-dubbed over the top of it in poorly translated and gibberish sounding English. The creator didn't figure out how to turn it off and it completely ruined her video.
If YT adds auto-dubbed language via AI, does that mean these videos get these tags when the original content didn't have any AI junk in it to begin with?
Upvote
37
(37
/
0)
I like catching history documentaries on YouTube, but lately the well has been poisoned by AI created junk, a lot of it with imagined facts. Tagging them may help.
Upvote
36
(36
/
0)
Yggdrasill1986
Smack-Fu Master, in training
Making it more obvious that a video is produced by AI or at least partially is a step in the right direction. But as faulty as the automated content detection system already is, I expect many Youtubers getting falsely flagged for that.
As a Youtuber myself, this is very annoying. I can remember at least half a dozens of times that one of my videos was falsely flagged for copyright infringement. Expanding that to AI is welcome, but I can already see how that will get used to deligitimate other Youtubers or Youtubers that will get flagged for no Ai content at all...
As a Youtuber myself, this is very annoying. I can remember at least half a dozens of times that one of my videos was falsely flagged for copyright infringement. Expanding that to AI is welcome, but I can already see how that will get used to deligitimate other Youtubers or Youtubers that will get flagged for no Ai content at all...
Upvote
21
(21
/
0)
ugh. i'm constantly trying to figure out if narrators are real or AI. as soon as i hear a voice with a narrow dynamic range, i start listening for the telltale rhythm and inflection issues. i'm sure i've dropped out on a few real speakers who just happen to have unusual speaking voices. but, that's the state of things...
Upvote
41
(41
/
0)
At least it makes scams easy to spot! If they can't be bothered to use a real narrator... or even a non-robotic text to speech... it's easy to instantly ignore.
Upvote
3
(3
/
0)
No need for buttons, rest assured that YouTube will provide "what's best for each user". I'm sure the YouTube algorithm already incorporates some sort of "user-likes-AI-video" score to hide or promote this type of content.
Upvote
3
(3
/
0)
One class of AI videos is very easy to identify: videos that begin with a prominent claim of not being made with AI. Every one of those is pure AI.
Upvote
-18
(1
/
-19)
Upvote
7
(8
/
-1)
It's amazing how normalized actual scam ads are now. Most high-profile sites literally pushing phishing attacks openly because they paid money to host those ads.
Upvote
14
(14
/
0)
I hope they apply this to music channels as well, there are music channels with hundreds of videos that couldn't be produced in the time frame listed (usually around 2024, if you're unlucky all the way back to 2022) and it makes it impossible to sort the wheat from the chaff when trying to explore a new music genre for background music, focus or relaxation.
There's a point where you can decide where that doesn't matter but then it gives the uploader passive revenue and a paper thin armor against deletion just shrugging and listening anyway (and many videos I have caught with random song titles but repeating songs)
But honestly, killing the plague of fake news first is the highest priority, including fake news that likes to label itself as entertainment fiction with poor fine print labelling and hope nobody notices their channel age is less than a year old in both cases.
There's a point where you can decide where that doesn't matter but then it gives the uploader passive revenue and a paper thin armor against deletion just shrugging and listening anyway (and many videos I have caught with random song titles but repeating songs)
But honestly, killing the plague of fake news first is the highest priority, including fake news that likes to label itself as entertainment fiction with poor fine print labelling and hope nobody notices their channel age is less than a year old in both cases.
Upvote
14
(14
/
0)
Merely a little tag in text outside the video? I think realistic "AI" slop should get a big "AI" watermark plastered over most of the video.
Upvote
15
(15
/
0)
Don't worry about that. Pretty soon running ad-blockers will be labelled 'ANTIFA' and 'anti-capitalist' and land you in the gulags as well due to newspeak, if the tech bro CEOs can hold the White House past 2028
https://www.whitehouse.gov/presiden...c-terrorism-and-organized-political-violence/
Upvote
18
(18
/
0)
A label does not solve the problem. How about a "show AI content" toggle that is off by default?
Upvote
22
(22
/
0)
I wish there was someway to reliably label an AI script. More than anything else, these annoy me. It usually take 30 seconds or so to realize it too.
Upvote
9
(9
/
0)
jonbob_newcastle
Ars Scholae Palatinae
There’s a YT channel that is pretty much an AI voice over a slideshow that talks about 40k lore. It used to have a fake David Attenborough voice and I thought that was pretty innocuous given that he’s been parodied and imitated so often, as well as the fact that no one would think it was really him. It was called Attenborough Lore I think, but they’ve changed the voice and the name now.
The new voice sounds a lot like Ian Richardson, another Brit with a pleasing voice, but it’s just not the same.
ed: added link to Ian Richardson interview.
Last edited:
Upvote
5
(5
/
0)
Not just voiceovers, but autodubs too. It's particularly obnoxious when a man's voice is used for multiple women.
I briefly considered getting a youtube subscription until realizing that you can't fully disable shorts. They've since demonstrated how good a call that was by adding algorithmic recommendations to the fricking subscriptions tab.
Upvote
7
(7
/
0)
RickRoyLeonPrisZhoraRacha
Ars Scholae Palatinae
I watch some Veteran Affairs videos that honor WW2/Korean/Vietnam/etc vets. In my recommendations are videos of "The tank that scared the Germans" or "The Germans captured this P51 and realized they lost the war!" .... and its all AI voice slop, with stills subjected to "film grain plugin". Its all for clicks, ads, views and analytics. No one should comment but they do.
And if you "stop recommending this channel" it will also be same slop under different name/channel.
Youtube will censor a historical demonstration of a WWI rifle firing but allow AI Slop to make ad revenue.
Upvote
14
(14
/
0)
Inflections are getting better. The best test is listen for a mispronounced word that no native speaker would get wrong.
Upvote
5
(5
/
0)
Or that a user can't tell the difference, which is probably most of their users.
Upvote
2
(2
/
0)
I am, but there are a few cases where it has been beneficial. I’m traveling overseas for work to a country where I don’t speak the language and have found a number of helpful videos that have what is clearly an AI-generated overdub in English. It sounds terrible, but it’s a lot better than not being able to get the information. That seems like a legit use case of AI, since translation services are way too expensive for a typical YouTube uploader to be able to afford.
But when I’m just watching some video about history or science that appears to be made by a native English speaker, I just have to ask “why?” It sounds so bad.
Upvote
4
(4
/
0)
Lets add the channels that just spam uploads of AI generated albums, particular of VGM covers.
Upvote
5
(5
/
0)
I also desperately want it for filtering out videos (or entire channels) of AI-generated music.
Upvote
4
(4
/
0)
What a crappy half-assed policy, why does AI animation get a pass from being called out as garbage slop?
Upvote
3
(3
/
0)
I was about to ask the same thing, why does it need to be photorealistic? I guess from the lying/propaganda standpoint realistic AI is more problematic than animated/artistic genAI but it doesn't help people who are trying to avoid it altogether, and that's before you go into other AI uses like voice/audio.
Upvote
9
(9
/
0)
I swear at least half the ads on YouTube are low-effort AI slop. I got an amusing one just yesterday—it was in the prolific genre of "stiff and obviously fake human robotically drones on about how your doctors are lying about diabetes because of Big Pharma and it's actually caused by 'a parasite that's suffocating your organs' and so you need to watch the scam video at the link below to learn this simple trick using cinnamon to cure diabetes forever".
This one had a new wrinkle claiming the pseudomedicine secret was deep Israeli knowledge (and that no one in Israel even knew what insulin shots were, hah), and so it featured a stereotypically Jewish guy... you know, the normal type of Israeli fellow who keeps a crucifix on his spacious kitchen counter.
The sad part is that the awful ad copy was clearly targeted toward seniors over 65, the kind of older people suspicious of institutions, confused by technology, and unable to tell when someone's blatantly lying to them. Which, as we know is a significant factor in our larger society's current, uh, situation.
Upvote
10
(10
/
0)